Daily Papers

In this manuscript we consider the problem of generalized linear estimation on Gaussian mixture data with labels given by a single-index model. Our first result is a sharp asymptotic expression for the test and training errors in the high-dimensional regime. Motivated by the recent stream of results on the Gaussian universality of the test and training errors in generalized linear estimation, we ask ourselves the question: "when is a single Gaussian enough to characterize the error?". Our formula allow us to give sharp answers to this question, both in the positive and negative directions. More precisely, we show that the sufficient conditions for Gaussian universality (or lack of thereof) crucially depend on the alignment between the target weights and the means and covariances of the mixture clusters, which we precisely quantify. In the particular case of least-squares interpolation, we prove a strong universality property of the training error, and show it follows a simple, closed-form expression. Finally, we apply our results to real datasets, clarifying some recent discussion in the literature about Gaussian universality of the errors in this context.

Continual Test-Time Adaptation (CTTA) aims to maintain model performance under evolving target domains by adapting online without labeled data. However, practical deployments often cannot retain the source dataset due to privacy or licensing constraints, and purely source-free CTTA methods tend to become unstable under long-term distribution shift, suffering from compounding self-training errors and catastrophic forgetting. We introduce DO-ALL (Distill Once, Adapt Life-Long), a plug-and-play framework that revisits source information in a compact and privacy-conscious form via Dataset Distillation (DD). Before deployment, DO-ALL performs DD to produce a small set of synthetic distilled anchors that summarize the source distribution. During adaptation, each target sample is matched with its most semantically aligned anchor, which provides a stable reference for various CTTA via source replay, representation alignment, and manifold-smoothing regularization. DO-ALL can be seamlessly integrated into existing CTTA algorithms, consistently improving long-term robustness across CIFAR100-C, ImageNet-C, and the CCC benchmark. This demonstrates the potential of leveraging DD to enable stable and continuous adaptation without retaining raw source data. The code is available at https://github.com/blue-531/DOALL.

Deep neural networks coupled with fast simulation and improved computation have led to recent successes in the field of reinforcement learning (RL). However, most current RL-based approaches fail to generalize since: (a) the gap between simulation and real world is so large that policy-learning approaches fail to transfer; (b) even if policy learning is done in real world, the data scarcity leads to failed generalization from training to test scenarios (e.g., due to different friction or object masses). Inspired from H-infinity control methods, we note that both modeling errors and differences in training and test scenarios can be viewed as extra forces/disturbances in the system. This paper proposes the idea of robust adversarial reinforcement learning (RARL), where we train an agent to operate in the presence of a destabilizing adversary that applies disturbance forces to the system. The jointly trained adversary is reinforced -- that is, it learns an optimal destabilization policy. We formulate the policy learning as a zero-sum, minimax objective function. Extensive experiments in multiple environments (InvertedPendulum, HalfCheetah, Swimmer, Hopper and Walker2d) conclusively demonstrate that our method (a) improves training stability; (b) is robust to differences in training/test conditions; and c) outperform the baseline even in the absence of the adversary.

Modern large language models demonstrate impressive capabilities in text generation and generalization. However, they often struggle with solving text editing tasks, particularly when it comes to correcting spelling errors and mistypings. In this paper, we present a methodology for generative spelling correction (SC), which was tested on English and Russian languages and potentially can be extended to any language with minor changes. Our research mainly focuses on exploring natural spelling errors and mistypings in texts and studying the ways those errors can be emulated in correct sentences to effectively enrich generative models' pre-train procedure. We investigate the impact of such emulations and the models' abilities across different text domains. In this work, we investigate two spelling corruption techniques: 1) first one mimics human behavior when making a mistake through leveraging statistics of errors from particular dataset and 2) second adds the most common spelling errors, keyboard miss clicks, and some heuristics within the texts. We conducted experiments employing various corruption strategies, models' architectures and sizes on the pre-training and fine-tuning stages and evaluated the models using single-domain and multi-domain test sets. As a practical outcome of our work, we introduce SAGE (Spell checking via Augmentation and Generative distribution Emulation) is a library for automatic generative SC that includes a family of pre-trained generative models and built-in augmentation algorithms.

A language model's ability to reflect on its own reasoning provides a key advantage for solving complex problems. While most recent research has focused on how this ability develops during reinforcement learning, we show that it actually begins to emerge much earlier - during the model's pre-training. To study this, we introduce deliberate errors into chains-of-thought and test whether the model can still arrive at the correct answer by recognizing and correcting these mistakes. By tracking performance across different stages of pre-training, we observe that this self-correcting ability appears early and improves steadily over time. For instance, an OLMo2-7B model pre-trained on 4 trillion tokens displays self-correction on our six self-reflection tasks.

Large language models (LLMs) have made significant strides in various natural language processing (NLP) tasks. Recent research shows that the moderately-sized LLMs often outperform their larger counterparts after task-specific fine-tuning. In this work, we delve into the process of adapting LLMs to specialize in document-level machine translation (DocMT) for a specific language pair. Firstly, we explore how prompt strategies affect downstream translation performance. Then, we conduct extensive experiments with two fine-tuning methods, three LLM backbones, and 18 translation tasks across nine language pairs. Our findings indicate that in some cases, these specialized models even surpass GPT-4 in translation performance, while they still significantly suffer from the off-target translation issue in others, even if they are exclusively fine-tuned on bilingual parallel documents. Furthermore, we provide an in-depth analysis of these LLMs tailored for DocMT, exploring aspects such as translation errors, discourse phenomena, training strategy, the scaling law of parallel documents, additional evaluation on recent test sets, and zero-shot crosslingual transfer. Our findings not only shed light on the strengths and limitations of LLM-based DocMT models but also provide a foundation for future research.

Backpropagation is the standard method for achieving state-of-the-art accuracy in neural network training, but it often imposes high memory costs and lacks biological plausibility. In this paper, we introduce the Mono-Forward algorithm, a purely local layerwise learning method inspired by Hinton's Forward-Forward framework. Unlike backpropagation, Mono-Forward optimizes each layer solely with locally available information, eliminating the reliance on global error signals. We evaluated Mono-Forward on multi-layer perceptrons and convolutional neural networks across multiple benchmarks, including MNIST, Fashion-MNIST, CIFAR-10, and CIFAR-100. The test results show that Mono-Forward consistently matches or surpasses the accuracy of backpropagation across all tasks, with significantly reduced and more even memory usage, better parallelizability, and a comparable convergence rate.

Process Reward Models (PRMs) provide step-level supervision that improves the reliability of reasoning in large language models. While PRMs have been extensively studied in text-based domains, their extension to Vision Language Models (VLMs) remains limited. Existing Vision-Language PRMs (VL-PRMs) rely on Monte Carlo Tree Search (MCTS) for data construction, which can often produce noisy supervision signals and limit generalization across tasks. In this work, we aim to elucidate the design space of VL-PRMs by exploring diverse strategies for dataset construction, training, and test-time scaling. First, we introduce a hybrid data synthesis framework that combines MCTS with judgments from a strong VLM, producing more accurate step-level labels. Second, we propose perception-focused supervision, enabling our PRM to explicitly detect errors at the visual grounding stage of reasoning. Third, we systematically evaluate multiple test-time scaling strategies, showing that our PRMs can reliably guide VLMs toward more accurate solutions. Our experiments covering five diverse multimodal benchmarks (MMMU, PuzzleVQA, AlgoPuzzleVQA, MathVista, and MathVision) reveal several key insights: (i) VL-PRMs when used as Outcome Reward Models (ORMs) during test-time scaling (TTS) can outperform VL-PRM guided process step selection, (ii) smaller VL-PRMs can match or even surpass larger ones in detecting process errors, (iii) VL-PRMs uncover latent reasoning abilities in stronger VLM backbones, (iv) perception-level supervision leads to significant gains in test-time scaling, and (v) TTS performance of different policies improve on advanced math reasoning datasets despite not training VL-PRMs on such datasets. We hope our work will motivate further research and support the advancement of VLMs.

Embodied LLMs endow robots with high-level task reasoning, but they cannot reflect on what went wrong or why, turning deployment into a sequence of independent trials where mistakes repeat rather than accumulate into experience. Drawing upon human reflective practitioners, we introduce Reflective Test-Time Planning, which integrates two modes of reflection: reflection-in-action, where the agent uses test-time scaling to generate and score multiple candidate actions using internal reflections before execution; and reflection-on-action, which uses test-time training to update both its internal reflection model and its action policy based on external reflections after execution. We also include retrospective reflection, allowing the agent to re-evaluate earlier decisions and perform model updates with hindsight for proper long-horizon credit assignment. Experiments on our newly-designed Long-Horizon Household benchmark and MuJoCo Cupboard Fitting benchmark show significant gains over baseline models, with ablative studies validating the complementary roles of reflection-in-action and reflection-on-action. Qualitative analyses, including real-robot trials, highlight behavioral correction through reflection.

Hallucinations are a type of output error produced by deep neural networks. While this has been studied in natural language processing, they have not been researched previously in automatic speech recognition. Here, we define hallucinations in ASR as transcriptions generated by a model that are semantically unrelated to the source utterance, yet still fluent and coherent. The similarity of hallucinations to probable natural language outputs of the model creates a danger of deception and impacts the credibility of the system. We show that commonly used metrics, such as word error rates, cannot differentiate between hallucinatory and non-hallucinatory models. To address this, we propose a perturbation-based method for assessing the susceptibility of an automatic speech recognition (ASR) model to hallucination at test time, which does not require access to the training dataset. We demonstrate that this method helps to distinguish between hallucinatory and non-hallucinatory models that have similar baseline word error rates. We further explore the relationship between the types of ASR errors and the types of dataset noise to determine what types of noise are most likely to create hallucinatory outputs. We devise a framework for identifying hallucinations by analysing their semantic connection with the ground truth and their fluency. Finally, we discover how to induce hallucinations with a random noise injection to the utterance.

Large language models (LLMs) demonstrate strong reasoning capabilities, but their performance often degrades under distribution shift. Existing test-time adaptation (TTA) methods rely on gradient-based updates that require white-box access and need substantial overhead, while training-free alternatives are either static or depend on external guidance. In this paper, we propose Training-Free Test-Time Contrastive Learning TF-TTCL, a training-free adaptation framework that enables a frozen LLM to improve online by distilling supervision from its own inference experiences. Specifically, TF-TTCL implements a dynamic "Explore-Reflect-Steer" loop through three core modules: 1) Semantic Query Augmentation first diversifies problem views via multi-agent role-playing to generate different reasoning trajectories; 2) Contrastive Experience Distillation then captures the semantic gap between superior and inferior trajectories, distilling them into explicit textual rules; and 3) Contextual Rule Retrieval finally activates these stored rules during inference to dynamically steer the frozen LLM toward robust reasoning patterns while avoiding observed errors. Extensive experiments on closed-ended reasoning tasks and open-ended evaluation tasks demonstrate that TF-TTCL consistently outperforms strong zero-shot baselines and representative TTA methods under online evaluation. Code is available at https://github.com/KevinSCUTer/TF-TTCL.

Generative Error Correction (GEC) has emerged as a powerful post-processing method to enhance the performance of Automatic Speech Recognition (ASR) systems. However, we show that GEC models struggle to generalize beyond the specific types of errors encountered during training, limiting their ability to correct new, unseen errors at test time, particularly in out-of-domain (OOD) scenarios. This phenomenon amplifies with named entities (NEs), where, in addition to insufficient contextual information or knowledge about the NEs, novel NEs keep emerging. To address these issues, we propose DARAG (Data- and Retrieval-Augmented Generative Error Correction), a novel approach designed to improve GEC for ASR in in-domain (ID) and OOD scenarios. We augment the GEC training dataset with synthetic data generated by prompting LLMs and text-to-speech models, thereby simulating additional errors from which the model can learn. For OOD scenarios, we simulate test-time errors from new domains similarly and in an unsupervised fashion. Additionally, to better handle named entities, we introduce retrieval-augmented correction by augmenting the input with entities retrieved from a database. Our approach is simple, scalable, and both domain- and language-agnostic. We experiment on multiple datasets and settings, showing that DARAG outperforms all our baselines, achieving 8\% -- 30\% relative WER improvements in ID and 10\% -- 33\% improvements in OOD settings.

ufakzeka-karar is an open Turkish decision model with 182,494,466 parameters. Given a Turkish text and questions of a fixed answer type (a choice, a level on an ordered scale, or yes or no), it returns a temperature-scaled probability for every option and an expected error that serves as a "not sure" signal, without generating text and in one CPU forward pass for up to ten options. Built on the lab's ufakzeka-1-base, its head scores each option blind to the others at shared positions, so the answer does not depend on option order. A sequential head trained with shuffled options was about as accurate but changed 2.3 to 2.8 percent of its answers when only the option order changed; REINFORCE lost 10.2 points (0.102) of macro F1 to cross-entropy. On the open set of HakemBench v1.0 (4,275 questions, 7 tracks) the released model ranks 7th of 16 rows with a composite of 0.660 (95% interval 0.642 to 0.677). Temperature scaling lowers calibration error (smooth ECE) on the development set but raises it on held-out support questions, from 0.027 to 0.045 for the first scored run, which never trained on them; the released model later trained on them, so its 0.036 to 0.064 is not an unseen-question test. The released model is the last of three runs scored on HakemBench, and its numbers are not blind. The second run's new training data was aimed at the first run's errors on the full test set in guardrails, moderation and customer support, and the released run was trained after the second run's guardrail results on the full test set were read, under a protocol fixed in writing before any of its data, code or runs. All its numbers come after these readings; its guardrail, moderation and customer support numbers carry the flag "shaped by reading the test results". With every model scored on the other four tracks only, its composite is 0.678, 6th of 16. Weights and code are under Apache-2.0.

The generalization performance of a machine learning algorithm such as a neural network depends in a non-trivial way on the structure of the data distribution. To analyze the influence of data structure on test loss dynamics, we study an exactly solveable model of stochastic gradient descent (SGD) on mean square loss which predicts test loss when training on features with arbitrary covariance structure. We solve the theory exactly for both Gaussian features and arbitrary features and we show that the simpler Gaussian model accurately predicts test loss of nonlinear random-feature models and deep neural networks trained with SGD on real datasets such as MNIST and CIFAR-10. We show that the optimal batch size at a fixed compute budget is typically small and depends on the feature correlation structure, demonstrating the computational benefits of SGD with small batch sizes. Lastly, we extend our theory to the more usual setting of stochastic gradient descent on a fixed subsampled training set, showing that both training and test error can be accurately predicted in our framework on real data.

We develop a novel approach to explain why AdaBoost is a successful classifier. By introducing a measure of the influence of the noise points (ION) in the training data for the binary classification problem, we prove that there is a strong connection between the ION and the test error. We further identify that the ION of AdaBoost decreases as the iteration number or the complexity of the base learners increases. We confirm that it is impossible to obtain a consistent classifier without deep trees as the base learners of AdaBoost in some complicated situations. We apply AdaBoost in portfolio management via empirical studies in the Chinese market, which corroborates our theoretical propositions.

Evaluating the performance of a well-trained GNN model on real-world graphs is a pivotal step for reliable GNN online deployment and serving. Due to a lack of test node labels and unknown potential training-test graph data distribution shifts, conventional model evaluation encounters limitations in calculating performance metrics (e.g., test error) and measuring graph data-level discrepancies, particularly when the training graph used for developing GNNs remains unobserved during test time. In this paper, we study a new research problem, online GNN evaluation, which aims to provide valuable insights into the well-trained GNNs's ability to effectively generalize to real-world unlabeled graphs under the test-time graph distribution shifts. Concretely, we develop an effective learning behavior discrepancy score, dubbed LeBeD, to estimate the test-time generalization errors of well-trained GNN models. Through a novel GNN re-training strategy with a parameter-free optimality criterion, the proposed LeBeD comprehensively integrates learning behavior discrepancies from both node prediction and structure reconstruction perspectives. This enables the effective evaluation of the well-trained GNNs' ability to capture test node semantics and structural representations, making it an expressive metric for estimating the generalization error in online GNN evaluation. Extensive experiments on real-world test graphs under diverse graph distribution shifts could verify the effectiveness of the proposed method, revealing its strong correlation with ground-truth test errors on various well-trained GNN models.

This paper focuses on over-parameterized deep neural networks (DNNs) with ReLU activation functions and proves that when the data distribution is well-separated, DNNs can achieve Bayes-optimal test error for classification while obtaining (nearly) zero-training error under the lazy training regime. For this purpose, we unify three interrelated concepts of overparameterization, benign overfitting, and the Lipschitz constant of DNNs. Our results indicate that interpolating with smoother functions leads to better generalization. Furthermore, we investigate the special case where interpolating smooth ground-truth functions is performed by DNNs under the Neural Tangent Kernel (NTK) regime for generalization. Our result demonstrates that the generalization error converges to a constant order that only depends on label noise and initialization noise, which theoretically verifies benign overfitting. Our analysis provides a tight lower bound on the normalized margin under non-smooth activation functions, as well as the minimum eigenvalue of NTK under high-dimensional settings, which has its own interest in learning theory.

Collecting audio-text pairs is expensive; however, it is much easier to access text-only data. Unless using shallow fusion, end-to-end automatic speech recognition (ASR) models require architecture modifications or additional training schemes to use text-only data. Inspired by recent advances in decoder-only language models (LMs), such as GPT-3 and PaLM adopted for speech-processing tasks, we propose using a decoder-only architecture for ASR with simple text augmentation. To provide audio information, encoder features compressed by CTC prediction are used as prompts for the decoder, which can be regarded as refining CTC prediction using the decoder-only model. Because the decoder architecture is the same as an autoregressive LM, it is simple to enhance the model by leveraging external text data with LM training. An experimental comparison using LibriSpeech and Switchboard shows that our proposed models with text augmentation training reduced word error rates from ordinary CTC by 0.3% and 1.4% on LibriSpeech test-clean and testother set, respectively, and 2.9% and 5.0% on Switchboard and CallHome. The proposed model had advantage on computational efficiency compared with conventional encoder-decoder ASR models with a similar parameter setup, and outperformed them on the LibriSpeech 100h and Switchboard training scenarios.

In our era of enormous neural networks, empirical progress has been driven by the philosophy that more is better. Recent deep learning practice has found repeatedly that larger model size, more data, and more computation (resulting in lower training loss) improves performance. In this paper, we give theoretical backing to these empirical observations by showing that these three properties hold in random feature (RF) regression, a class of models equivalent to shallow networks with only the last layer trained. Concretely, we first show that the test risk of RF regression decreases monotonically with both the number of features and the number of samples, provided the ridge penalty is tuned optimally. In particular, this implies that infinite width RF architectures are preferable to those of any finite width. We then proceed to demonstrate that, for a large class of tasks characterized by powerlaw eigenstructure, training to near-zero training loss is obligatory: near-optimal performance can only be achieved when the training error is much smaller than the test error. Grounding our theory in real-world data, we find empirically that standard computer vision tasks with convolutional neural tangent kernels clearly fall into this class. Taken together, our results tell a simple, testable story of the benefits of overparameterization, overfitting, and more data in random feature models.

Recent advances in large language models (LLMs) have improved their performance on coding benchmarks. However, improvement is plateauing due to the exhaustion of readily available high-quality data. Prior work has shown the potential of synthetic self-instruct data, but naively training on a model's own outputs can cause error accumulation, especially in coding tasks, where generalization may collapse due to overly simple or erroneous training data, highlighting the need for rigorous quality checks on synthetic data. In this work, we explore an effective approach whereby the model itself verifies the correctness of its own data. We thus propose Sol-Ver, a self-play solver-verifier framework that jointly improves a single model's code and test generation capacity. By iteratively refining code (LLM-as-a-solver) and tests (LLM-as-a-verifier) together, we boost both capabilities without relying on human annotations or larger teacher models. Experiments with the Llama 3.1 8B model demonstrate substantial performance enhancements, achieving average relative improvements of 19.63% in code generation and 17.49% in test generation on MBPP and LiveCodeBench.

Data-driven surrogate models are increasingly adopted to accelerate vehicle design. However, open-source multi-fidelity datasets and empirical guidelines linking dataset size to model performance remain limited. This study investigates the relationship between training data size and prediction accuracy for a graph neural network (GNN) based surrogate model for aerodynamic field prediction. We release an open-source, multi-fidelity aerodynamic dataset for double-delta wings, comprising 2448 flow snapshots across 272 geometries evaluated at angles of attack from 11 (degree) to 19 (degree) at Ma=0.3 using both Vortex Lattice Method (VLM) and Reynolds-Averaged Navier-Stokes (RANS) solvers. The geometries are generated using a nested Saltelli sampling scheme to support future dataset expansion and variance-based sensitivity analysis. Using this dataset, we conduct a preliminary empirical scaling study of the MF-VortexNet surrogate by constructing six training datasets with sizes ranging from 40 to 1280 snapshots and training models with 0.1 to 2.4 million parameters under a fixed training budget. We find that the test error decreases with data size with a power-law exponent of -0.6122, indicating efficient data utilization. Based on this scaling law, we estimate that the optimal sampling density is approximately eight samples per dimension in a d-dimensional design space. The results also suggest improved data utilization efficiency for larger surrogate models, implying a potential trade-off between dataset generation cost and model training budget.

With the growing adoption of Large Language Model (LLM) agents in persistent, real-world roles, they naturally encounter continuous streams of tasks and inevitable failures. A key limitation, however, is their inability to systematically learn from these mistakes, forcing them to repeat identical errors in similar contexts. Unlike prior training-free methods that primarily store raw instance-level experience or focus on retrieving successful trajectories, we propose Mistake Notebook Learning (MNL), a novel memory framework that enables agents to self-curate generalizable guidance from batch-clustered failures. This mechanism allows agents to distill shared error patterns into structured "mistake notes," updating an external memory only when batch performance improves to ensure stability. To further amplify adaptability, we integrate MNL with test-time scaling, leveraging aggregated failure patterns to actively steer the search process away from known pitfalls. Experiments on mathematical reasoning, Text-to-SQL, and interactive agent benchmarks show that MNL achieves competitive performance compared to existing memory mechanisms and in-context methods in both effectiveness and efficiency. These findings position structured mistake abstraction as a critical lever for robust agent evolution, enabling continuous improvement without the cost of parameter updates. The code is available at https://github.com/Bairong-Xdynamics/MistakeNotebookLearning/tree/main.

Conformer, combining convolution and self-attention sequentially to capture both local and global information, has shown remarkable performance and is currently regarded as the state-of-the-art for automatic speech recognition (ASR). Several other studies have explored integrating convolution and self-attention but they have not managed to match Conformer's performance. The recently introduced Branchformer achieves comparable performance to Conformer by using dedicated branches of convolution and self-attention and merging local and global context from each branch. In this paper, we propose E-Branchformer, which enhances Branchformer by applying an effective merging method and stacking additional point-wise modules. E-Branchformer sets new state-of-the-art word error rates (WERs) 1.81% and 3.65% on LibriSpeech test-clean and test-other sets without using any external training data.

The evaluation of Handwritten Text Recognition (HTR) models during their development is straightforward: because HTR is a supervised problem, the usual data split into training, validation, and test data sets allows the evaluation of models in terms of accuracy or error rates. However, the evaluation process becomes tricky as soon as we switch from development to application. A compilation of a new (and forcibly smaller) ground truth (GT) from a sample of the data that we want to apply the model on and the subsequent evaluation of models thereon only provides hints about the quality of the recognised text, as do confidence scores (if available) the models return. Moreover, if we have several models at hand, we face a model selection problem since we want to obtain the best possible result during the application phase. This calls for GT-free metrics to select the best model, which is why we (re-)introduce and compare different metrics, from simple, lexicon-based to more elaborate ones using standard language models and masked language models (MLM). We show that MLM-based evaluation can compete with lexicon-based methods, with the advantage that large and multilingual transformers are readily available, thus making compiling lexical resources for other metrics superfluous.

Test-time scaling via sequential revision has emerged as a powerful paradigm for enhancing Large Language Model (LLM) reasoning. However, standard post-training methods primarily optimize single-shot objectives, creating a fundamental misalignment with multi-step inference dynamics. While recent work treats this as multi-turn reinforcement learning (RL), conventional approaches optimize over the multi-step trajectories directly, failing to further exploit the high-quality mistakes in intermediate steps that model can learn from correcting them. We propose a two-stage iterative framework that alternates between online data/prompt augmentation and policy optimization. By converting the intermediate steps (``near-miss'' answers) in the successful recovery trajectories into decoupled revision and verification prompts, our approach concentrates training on both effective answer transformation and error identification. This approach enables efficient off-policy data generation and reduces the computational overhead of long-horizon sampling compared to standard multi-turn RL. On LiveCodeBench, using publicly available test cases as feedback, we observe gains of +6.5 points over the RL baseline and +4.0 points over standard multi-turn training. Beyond coding, our approach matches the previously reported SOTA result on circle packing while using the smallest base model (4B) and far fewer rollouts than the much larger evolutionary search systems. Math results under ground-truth verification further confirm improved correction ability. It also generalizes to out-of-distribution constraint-satisfaction puzzles such as n\_queens and mini\_sudoku, where correctness is defined entirely by problem constraints. Code is available at https://github.com/yxliu02/REVES.git.

This paper presents a novel Dialectal Sound and Vowelization Recovery framework, designed to recognize borrowed and dialectal sounds within phonologically diverse and dialect-rich languages, that extends beyond its standard orthographic sound sets. The proposed framework utilized a quantized sequence of input with(out) continuous pretrained self-supervised representation. We show the efficacy of the pipeline using limited data for Arabic, a dialect-rich language containing more than 22 major dialects. Phonetically correct transcribed speech resources for dialectal Arabic are scarce. Therefore, we introduce ArabVoice15, a first-of-its-kind, curated test set featuring 5 hours of dialectal speech across 15 Arab countries, with phonetically accurate transcriptions, including borrowed and dialect-specific sounds. We described in detail the annotation guideline along with the analysis of the dialectal confusion pairs. Our extensive evaluation includes both subjective -- human perception tests and objective measures. Our empirical results, reported with three test sets, show that with only one and half hours of training data, our model improve character error rate by ~ 7\% in ArabVoice15 compared to the baseline.

We propose Stable Video Infinity (SVI) that is able to generate infinite-length videos with high temporal consistency, plausible scene transitions, and controllable streaming storylines. While existing long-video methods attempt to mitigate accumulated errors via handcrafted anti-drifting (e.g., modified noise scheduler, frame anchoring), they remain limited to single-prompt extrapolation, producing homogeneous scenes with repetitive motions. We identify that the fundamental challenge extends beyond error accumulation to a critical discrepancy between the training assumption (seeing clean data) and the test-time autoregressive reality (conditioning on self-generated, error-prone outputs). To bridge this hypothesis gap, SVI incorporates Error-Recycling Fine-Tuning, a new type of efficient training that recycles the Diffusion Transformer (DiT)'s self-generated errors into supervisory prompts, thereby encouraging DiT to actively identify and correct its own errors. This is achieved by injecting, collecting, and banking errors through closed-loop recycling, autoregressively learning from error-injected feedback. Specifically, we (i) inject historical errors made by DiT to intervene on clean inputs, simulating error-accumulated trajectories in flow matching; (ii) efficiently approximate predictions with one-step bidirectional integration and calculate errors with residuals; (iii) dynamically bank errors into replay memory across discretized timesteps, which are resampled for new input. SVI is able to scale videos from seconds to infinite durations with no additional inference cost, while remaining compatible with diverse conditions (e.g., audio, skeleton, and text streams). We evaluate SVI on three benchmarks, including consistent, creative, and conditional settings, thoroughly verifying its versatility and state-of-the-art role.

Training Deep Neural Networks is complicated by the fact that the distribution of each layer's inputs changes during training, as the parameters of the previous layers change. This slows down the training by requiring lower learning rates and careful parameter initialization, and makes it notoriously hard to train models with saturating nonlinearities. We refer to this phenomenon as internal covariate shift, and address the problem by normalizing layer inputs. Our method draws its strength from making normalization a part of the model architecture and performing the normalization for each training mini-batch. Batch Normalization allows us to use much higher learning rates and be less careful about initialization. It also acts as a regularizer, in some cases eliminating the need for Dropout. Applied to a state-of-the-art image classification model, Batch Normalization achieves the same accuracy with 14 times fewer training steps, and beats the original model by a significant margin. Using an ensemble of batch-normalized networks, we improve upon the best published result on ImageNet classification: reaching 4.9% top-5 validation error (and 4.8% test error), exceeding the accuracy of human raters.

Test-Time Training (TTT) models context dependencies by adapting part of the model's weights (referred to as fast weights) during inference. This fast weight, akin to recurrent states in RNNs, stores temporary memories of past tokens in the current sequence. Existing TTT methods struggled to show effectiveness in handling long-context data, due to their inefficiency on modern GPUs. The TTT layers in many of these approaches operate with extremely low FLOPs utilization (often <5%) because they deliberately apply small online minibatch sizes (e.g., updating fast weights every 16 or 64 tokens). Moreover, a small minibatch implies fine-grained block-wise causal dependencies in the data, unsuitable for data beyond 1D ordered sequences, like sets or N-dimensional grids such as images or videos. In contrast, we pursue the opposite direction by using an extremely large chunk update, ranging from 2K to 1M tokens across tasks of varying modalities, which we refer to as Large Chunk Test-Time Training (LaCT). It improves hardware utilization by orders of magnitude, and more importantly, facilitates scaling of nonlinear state size (up to 40% of model parameters), hence substantially improving state capacity, all without requiring cumbersome and error-prone kernel implementations. It also allows easy integration of sophisticated optimizers, e.g. Muon for online updates. We validate our approach across diverse modalities and tasks, including novel view synthesis with image set, language models, and auto-regressive video diffusion. Our approach can scale up to 14B-parameter AR video diffusion model on sequences up to 56K tokens. In our longest sequence experiment, we perform novel view synthesis with 1 million context length. We hope this work will inspire and accelerate new research in the field of long-context modeling and test-time training. Website: https://tianyuanzhang.com/projects/ttt-done-right

Achieving truly adaptive embodied intelligence requires agents that learn not just by imitating static demonstrations, but by continuously improving through environmental interaction, which is akin to how humans master skills through practice. Vision-Language-Action (VLA) models have advanced robotic manipulation by leveraging large language models, yet remain fundamentally limited by Supervised Finetuning (SFT): requiring hundreds of demonstrations per task, rigidly memorizing trajectories, and failing to adapt when deployment conditions deviate from training. We introduce EVOLVE-VLA, a test-time training framework enabling VLAs to continuously adapt through environment interaction with minimal or zero task-specific demonstrations. The key technical challenge is replacing oracle reward signals (unavailable at test time) with autonomous feedback. We address this through a learned progress estimator providing dense feedback, and critically, we design our framework to ``tame'' this inherently noisy signal via two mechanisms: (1) an accumulative progress estimation mechanism smoothing noisy point-wise estimates, and (2) a progressive horizon extension strategy enabling gradual policy evolution. EVOLVE-VLA achieves substantial gains: +8.6\% on long-horizon tasks, +22.0\% in 1-shot learning, and enables cross-task generalization -- achieving 20.8\% success on unseen tasks without task-specific demonstrations training (vs. 0\% for pure SFT). Qualitative analysis reveals emergent capabilities absent in demonstrations, including error recovery and novel strategies. This work represents a critical step toward VLAs that truly learn and adapt, moving beyond static imitation toward continuous self-improvements.

World models enable agents to plan by predicting future states of the environment, but their predictions can become unreliable when test-time dynamics differ from those seen during training. We present JEPA-TTT, which adapts the latent dynamics predictor of a pretrained action-conditioned Joint-Embedding Predictive Architecture world model throughout test time. Self-supervised updates accumulate across episodes, while the visual encoder and reward head remain fixed, preserving the pretrained representation and task objective. Planning requires neither a goal image nor online environment reward. JEPA-TTT uses dense replay, which forms prediction windows at every temporal offset, retains them in a growing buffer, and samples minibatches from that buffer for predictor updates. Across eight dynamics shifts in four continuous-control environments, JEPA-TTT improves planning on every shift. After 500 test-time episodes, it reduces autoregressive latent prediction error by 83% on average and improves planning performance by 153% over the frozen JEPA world model. These results show that persistent self-supervised test-time training can adapt a pretrained latent world model under changed dynamics.

While test-time scaling has enabled large language models to solve highly difficult tasks, state-of-the-art results come at exorbitant compute costs. These inefficiencies can be attributed to the miscalibration of post-trained language models, and the lack of calibration in popular sampling techniques. Here, we present Online Reasoning Calibration (ORCA), a framework for calibrating the sampling process that draws upon conformal prediction and test-time training. Specifically, we introduce a meta-learning procedure that updates the calibration module for each input. This allows us to provide valid confidence estimates under distributional shift, e.g. in thought patterns that occur across different stages of reasoning, or in prompt distributions between model development and deployment. ORCA not only provides theoretical guarantees on conformal risks, but also empirically shows higher efficiency and generalization across different reasoning tasks. At risk level δ=0.1, ORCA improves Qwen2.5-32B efficiency on in-distribution tasks with savings up to 47.5% with supervised labels and 40.7% with self-consistency labels. Under zero-shot out-of-domain settings, it improves MATH-500 savings from 24.8% of the static calibration baseline to 67.0% while maintaining a low empirical error rate, and the same trend holds across model families and downstream benchmarks. Our code is publicly available at https://github.com/wzekai99/ORCA.

Joint punctuated and normalized automatic speech recognition (ASR) aims at outputing transcripts with and without punctuation and casing. This task remains challenging due to the lack of paired speech and punctuated text data in most ASR corpora. We propose two approaches to train an end-to-end joint punctuated and normalized ASR system using limited punctuated data. The first approach uses a language model to convert normalized training transcripts into punctuated transcripts. This achieves a better performance on out-of-domain test data, with up to 17% relative Punctuation-Case-aware Word Error Rate (PC-WER) reduction. The second approach uses a single decoder conditioned on the type of output. This yields a 42% relative PC-WER reduction compared to Whisper-base and a 4% relative (normalized) WER reduction compared to the normalized output of a punctuated-only model. Additionally, our proposed model demonstrates the feasibility of a joint ASR system using as little as 5% punctuated training data with a moderate (2.42% absolute) PC-WER increase.

As Large Multimodal Models (LMMs) scale up and reinforcement learning (RL) methods mature, LMMs have made notable progress in complex reasoning and decision making. Yet training still relies on static data and fixed recipes, making it difficult to diagnose capability blind spots or provide dynamic, targeted reinforcement. Motivated by findings that test driven error exposure and feedback based correction outperform repetitive practice, we propose Diagnostic-driven Progressive Evolution (DPE), a spiral loop where diagnosis steers data generation and reinforcement, and each iteration re-diagnoses the updated model to drive the next round of targeted improvement. DPE has two key components. First, multiple agents annotate and quality control massive unlabeled multimodal data, using tools such as web search and image editing to produce diverse, realistic samples. Second, DPE attributes failures to specific weaknesses, dynamically adjusts the data mixture, and guides agents to generate weakness focused data for targeted reinforcement. Experiments on Qwen3-VL-8B-Instruct and Qwen2.5-VL-7B-Instruct show stable, continual gains across eleven benchmarks, indicating DPE as a scalable paradigm for continual LMM training under open task distributions. Our code, models, and data are publicly available at https://github.com/hongruijia/DPE.

Several machine learning models, including neural networks, consistently misclassify adversarial examples---inputs formed by applying small but intentionally worst-case perturbations to examples from the dataset, such that the perturbed input results in the model outputting an incorrect answer with high confidence. Early attempts at explaining this phenomenon focused on nonlinearity and overfitting. We argue instead that the primary cause of neural networks' vulnerability to adversarial perturbation is their linear nature. This explanation is supported by new quantitative results while giving the first explanation of the most intriguing fact about them: their generalization across architectures and training sets. Moreover, this view yields a simple and fast method of generating adversarial examples. Using this approach to provide examples for adversarial training, we reduce the test set error of a maxout network on the MNIST dataset.

We estimate ([M/H], [alpha/M]) for 48 million giants and dwarfs in low-dust extinction regions from the Gaia DR3 XP spectra by using tree-based machine-learning models trained on APOGEE DR17 and metal-poor star sample from Li et al. The root mean square error of our estimation is 0.0890 dex for [M/H] and 0.0436 dex for [alpha/M], when we evaluate our models on the test data that are not used in training the models. Because the training data is dominated by giants, our estimation is most reliable for giants. The high-[alpha/M] stars and low-[alpha/M] stars selected by our ([M/H], [alpha/M]) show different kinematical properties for giants and low-temperature dwarfs. We further investigate how our machine-learning models extract information on ([M/H], [alpha/M]). Intriguingly, we find that our models seem to extract information on [alpha/M] from Na D lines (589 nm) and Mg I line (516 nm). This result is understandable given the observed correlation between Na and Mg abundances in the literature. The catalog of ([M/H], [alpha/M]) as well as their associated uncertainties are publicly available online.

Recent success in deep learning has partially been driven by training increasingly overparametrized networks on ever larger datasets. It is therefore natural to ask: how much of the data is superfluous, which examples are important for generalization, and how do we find them? In this work, we make the striking observation that, in standard vision datasets, simple scores averaged over several weight initializations can be used to identify important examples very early in training. We propose two such scores -- the Gradient Normed (GraNd) and the Error L2-Norm (EL2N) scores -- and demonstrate their efficacy on a range of architectures and datasets by pruning significant fractions of training data without sacrificing test accuracy. In fact, using EL2N scores calculated a few epochs into training, we can prune half of the CIFAR10 training set while slightly improving test accuracy. Furthermore, for a given dataset, EL2N scores from one architecture or hyperparameter configuration generalize to other configurations. Compared to recent work that prunes data by discarding examples that are rarely forgotten over the course of training, our scores use only local information early in training. We also use our scores to detect noisy examples and study training dynamics through the lens of important examples -- we investigate how the data distribution shapes the loss surface and identify subspaces of the model's data representation that are relatively stable over training.

We present Noisy Student Training, a semi-supervised learning approach that works well even when labeled data is abundant. Noisy Student Training achieves 88.4% top-1 accuracy on ImageNet, which is 2.0% better than the state-of-the-art model that requires 3.5B weakly labeled Instagram images. On robustness test sets, it improves ImageNet-A top-1 accuracy from 61.0% to 83.7%, reduces ImageNet-C mean corruption error from 45.7 to 28.3, and reduces ImageNet-P mean flip rate from 27.8 to 12.2. Noisy Student Training extends the idea of self-training and distillation with the use of equal-or-larger student models and noise added to the student during learning. On ImageNet, we first train an EfficientNet model on labeled images and use it as a teacher to generate pseudo labels for 300M unlabeled images. We then train a larger EfficientNet as a student model on the combination of labeled and pseudo labeled images. We iterate this process by putting back the student as the teacher. During the learning of the student, we inject noise such as dropout, stochastic depth, and data augmentation via RandAugment to the student so that the student generalizes better than the teacher. Models are available at https://github.com/tensorflow/tpu/tree/master/models/official/efficientnet. Code is available at https://github.com/google-research/noisystudent.

With the proliferation of Audio Language Model (ALM) based deepfake audio, there is an urgent need for effective detection methods. Unlike traditional deepfake audio generation, which often involves multi-step processes culminating in vocoder usage, ALM directly utilizes neural codec methods to decode discrete codes into audio. Moreover, driven by large-scale data, ALMs exhibit remarkable robustness and versatility, posing a significant challenge to current audio deepfake detection (ADD) models. To effectively detect ALM-based deepfake audio, we focus on the mechanism of the ALM-based audio generation method, the conversion from neural codec to waveform. We initially construct the Codecfake dataset, an open-source large-scale dataset, including two languages, millions of audio samples, and various test conditions, tailored for ALM-based audio detection. Additionally, to achieve universal detection of deepfake audio and tackle domain ascent bias issue of original SAM, we propose the CSAM strategy to learn a domain balanced and generalized minima. Experiment results demonstrate that co-training on Codecfake dataset and vocoded dataset with CSAM strategy yield the lowest average Equal Error Rate (EER) of 0.616% across all test conditions compared to baseline models.

Non-contact remote photoplethysmography (rPPG) technology enables heart rate measurement from facial videos. However, existing network models still face challenges in accu racy, robustness, and generalization capability under complex scenarios. This paper proposes an end-to-end rPPG extraction network that employs 3D convolutional neural networks to reconstruct accurate rPPG signals from raw facial videos. We introduce a differential frame fusion module that integrates differential frames with original frames, enabling frame-level representations to capture blood volume pulse (BVP) variations. Additionally, we incorporate Temporal Shift Module (TSM) with self-attention mechanisms, which effectively enhance rPPG features with minimal computational overhead. Furthermore, we propose a novel dynamic hybrid loss function that provides stronger supervision for the network, effectively mitigating over fitting. Comprehensive experiments were conducted on not only the PURE and UBFC-rPPG datasets but also the challenging MMPD dataset under complex scenarios, involving both intra dataset and cross-dataset evaluations, which demonstrate the superior robustness and generalization capability of our network. Specifically, after training on PURE, our model achieved a mean absolute error (MAE) of 7.58 on the MMPD test set, outperforming the state-of-the-art models.

We present a variety of new architectural features and training procedures that we apply to the generative adversarial networks (GANs) framework. We focus on two applications of GANs: semi-supervised learning, and the generation of images that humans find visually realistic. Unlike most work on generative models, our primary goal is not to train a model that assigns high likelihood to test data, nor do we require the model to be able to learn well without using any labels. Using our new techniques, we achieve state-of-the-art results in semi-supervised classification on MNIST, CIFAR-10 and SVHN. The generated images are of high quality as confirmed by a visual Turing test: our model generates MNIST samples that humans cannot distinguish from real data, and CIFAR-10 samples that yield a human error rate of 21.3%. We also present ImageNet samples with unprecedented resolution and show that our methods enable the model to learn recognizable features of ImageNet classes.

Sequence-to-sequence attention-based models on subword units allow simple open-vocabulary end-to-end speech recognition. In this work, we show that such models can achieve competitive results on the Switchboard 300h and LibriSpeech 1000h tasks. In particular, we report the state-of-the-art word error rates (WER) of 3.54% on the dev-clean and 3.82% on the test-clean evaluation subsets of LibriSpeech. We introduce a new pretraining scheme by starting with a high time reduction factor and lowering it during training, which is crucial both for convergence and final performance. In some experiments, we also use an auxiliary CTC loss function to help the convergence. In addition, we train long short-term memory (LSTM) language models on subword units. By shallow fusion, we report up to 27% relative improvements in WER over the attention baseline without a language model.

In this paper, we introduce a new process prejudge strategy in LLM reasoning to demonstrate that bootstrapping with process prejudge allows the LLM to adaptively anticipate the errors encountered when advancing the subsequent reasoning steps, similar to people sometimes pausing to think about what mistakes may occur and how to avoid them, rather than relying solely on trial and error. Specifically, we define a prejudge node in the rationale, which represents a reasoning step, with at least one step that follows the prejudge node that has no paths toward the correct answer. To synthesize the prejudge reasoning process, we present an automated reasoning framework with a dynamic tree-searching strategy. This framework requires only one LLM to perform answer judging, response critiquing, prejudge generation, and thought completion. Furthermore, we develop a two-phase training mechanism with supervised fine-tuning (SFT) and reinforcement learning (RL) to further enhance the reasoning capabilities of LLMs. Experimental results from competition-level complex reasoning demonstrate that our method can teach the model to prejudge before thinking and significantly enhance the reasoning ability of LLMs. Code and data is released at https://github.com/wjn1996/Prejudge-Before-Think.

The prevalence of automatic speech recognition (ASR) systems in spoken language applications has increased significantly in recent years. Notably, many African languages lack sufficient linguistic resources to support the robustness of these systems. This paper focuses on the development of an end-to-end speech recognition system customized for Nigerian Pidgin English. We investigated and evaluated different pretrained state-of-the-art architectures on a new dataset. Our empirical results demonstrate a notable performance of the variant Wav2Vec2 XLSR-53 on our dataset, achieving a word error rate (WER) of 29.6% on the test set, surpassing other architectures such as NEMO QUARTZNET and Wav2Vec2.0 BASE-100H in quantitative assessments. Additionally, we demonstrate that pretrained state-of-the-art architectures do not work well out-of-the-box. We performed zero-shot evaluation using XLSR-English as the baseline, chosen for its similarity to Nigerian Pidgin. This yielded a higher WER of 73.7%. By adapting this architecture to nuances represented in our dataset, we reduce error by 59.84%. Our dataset comprises 4,288 recorded utterances from 10 native speakers, partitioned into training, validation, and test sets. This study underscores the potential for improving ASR systems for under-resourced languages like Nigerian Pidgin English, contributing to greater inclusion in speech technology applications. We publicly release our unique parallel dataset (speech-to-text) on Nigerian Pidgin, as well as the model weights on Hugging Face. Our code would be made available to foster future research from the community.

Very deep convolutional networks with hundreds of layers have led to significant reductions in error on competitive benchmarks. Although the unmatched expressiveness of the many layers can be highly desirable at test time, training very deep networks comes with its own set of challenges. The gradients can vanish, the forward flow often diminishes, and the training time can be painfully slow. To address these problems, we propose stochastic depth, a training procedure that enables the seemingly contradictory setup to train short networks and use deep networks at test time. We start with very deep networks but during training, for each mini-batch, randomly drop a subset of layers and bypass them with the identity function. This simple approach complements the recent success of residual networks. It reduces training time substantially and improves the test error significantly on almost all data sets that we used for evaluation. With stochastic depth we can increase the depth of residual networks even beyond 1200 layers and still yield meaningful improvements in test error (4.91% on CIFAR-10).

Mobile agents have achieved promising results on clean online benchmarks such as AndroidWorld, yet their performance often degrades sharply in real-world deployment due to environmental variations and imperfect interface conditions. In this work, we introduce AndroidReality, a perturbation-based framework for evaluating and improving the robustness of mobile agents. Through a Markov Decision Process (MDP) perspective, we organize real-world interface variability into a principled taxonomy of perturbations along three axes: state, transition, and action. Guided by this taxonomy, we build a perturbed mobile benchmark on top of AndroidWorld with realistic and controllable perturbation injections, enabling systematic robustness evaluation of mobile agents. Our evaluation reveals substantial robustness gaps and four recurring error categories, motivating a simple training-free Test-Time Introspective Recovery (TTIR) mechanism that mitigates these failures on both perturbed and clean settings. Together, these results position robustness as a missing dimension in mobile agent evaluation and establish benchmark perturbation as an effective tool for both stress testing and surfacing latent weaknesses of mobile agents.

Quantization is widely used to reduce the computational and memory demands of neural-network inference. In recurrent networks, however, the quantized state is stored and returned at the next time step, so the rule used to store that state can alter subsequent computations. Here, we introduce recurrent-state write-back to denote this rule and isolate its effect in a compact GRU encoder--decoder for fluorescence lifetime imaging, a molecular imaging modality used in quantitative biological imaging. A central task is estimating two lifetime parameters, the short-lived component τ1 and the long-lived component τ2, from high-noise time-resolved fluorescence signals. Holding the trained model fixed, replacing continuous state propagation with deterministic 4-bit state storage increases estimation errors for τ1 and τ2 by approximately 70x and 300x, respectively. Failure occurs when repeated small updates remain below the write threshold, leaving the stored state nearly fixed while the network continues to propose change. Error feedback, residual memory, and direction memory carry information from these suppressed updates across time and recover accuracy without retraining. Precision sweeps show that increasing state precision can worsen a fixed recurrent solution, while matched training shows that compatibility with the state interface can be learned. To test whether this behavior extends beyond the GRU, we repeat the post-training intervention in an independently trained LSTM, where coarse write-back reproduces the failure, error feedback restores accuracy, and state-specific interventions reveal greater sensitivity of the cell state than the hidden state. Our results establish recurrent-state write-back as a key determinant of low-precision recurrent dynamics and identify the state-storage interface as a central design consideration for quantized recurrent inference.

Neural architecture search (NAS) has shown encouraging results in automating the architecture design. Recently, DARTS relaxes the search process with a differentiable formulation that leverages weight-sharing and SGD where all candidate operations are trained simultaneously. Our empirical results show that such procedure results in the co-adaption problem and Matthew Effect: operations with fewer parameters would be trained maturely earlier. This causes two problems: firstly, the operations with more parameters may never have the chance to express the desired function since those with less have already done the job; secondly, the system will punish those underperforming operations by lowering their architecture parameter, and they will get smaller loss gradients, which causes the Matthew Effect. In this paper, we systematically study these problems and propose a novel grouped operation dropout algorithm named DropNAS to fix the problems with DARTS. Extensive experiments demonstrate that DropNAS solves the above issues and achieves promising performance. Specifically, DropNAS achieves 2.26% test error on CIFAR-10, 16.39% on CIFAR-100 and 23.4% on ImageNet (with the same training hyperparameters as DARTS for a fair comparison). It is also observed that DropNAS is robust across variants of the DARTS search space. Code is available at https://github.com/wiljohnhong/DropNAS.

An accurate solar wind speed model is important for space weather predictions, catastrophic event warnings, and other issues concerning solar wind - magnetosphere interaction. In this work, we construct a model based on convolutional neural network (CNN) and Potential Field Source Surface (PFSS) magnetograms, considering a solar wind source surface of R_{rm SS}=2.5R_odot, aiming to predict the solar wind speed at the Lagrange 1 (L1) point of the Sun-Earth system. The input of our model consists of four Potential Field Source Surface (PFSS) magnetograms at R_{rm SS}, which are 7, 6, 5, and 4 days before the target epoch. Reduced magnetograms are used to promote the model's efficiency. We use the Global Oscillation Network Group (GONG) photospheric magnetograms and the potential field extrapolation model to generate PFSS magnetograms at the source surface. The model provides predictions of the continuous test dataset with an averaged correlation coefficient (CC) of 0.52 and a root mean square error (RMSE) of 80.8 km/s in an eight-fold validation training scheme with the time resolution of the data as small as one hour. The model also has the potential to forecast high speed streams of the solar wind, which can be quantified with a general threat score of 0.39.

Large Reasoning Models (LRMs) exhibit backtracking and self-verification mechanisms that enable them to revise intermediate steps and reach correct solutions, yielding strong performance on complex logical benchmarks. We hypothesize that such behaviors are beneficial only when the model has sufficiently strong ``critique'' ability to detect its own mistakes. This work systematically investigates how current LRMs recover from errors by inserting arithmetic mistakes in their intermediate reasoning steps. Notably, we discover a peculiar yet important phenomenon: despite the error propagating throughout the entire chain-of-thought (CoT) without any verbalized correction, the model still reaches the correct final answer after the thinking process finishes. This recovery implies the existence of an internal mechanism helping the model to detect errors and trigger self-correction, which we refer to as the hidden critique ability. Building on feature space analysis, we identify a highly interpretable critique vector representing this behavior. Extensive experiments across multiple model scales and families demonstrate that steering latent representations with this vector improves the model's error detection capability and enhances the performance of test-time scaling at no extra training cost. Our findings provide a valuable understanding of LRMs' critique behavior, suggesting a promising direction to control and improve their self-verification mechanism. Our code is available at: https://github.com/mail-research/lrm-critique-vectors.

Reinforcement learning (RL) is credited with teaching reasoning models to write {wait} before reconsidering a step, yet large RL accuracy gains do not require the word. This view can confuse what RL discovers with what a model inherits from its training data. We trace the word through open models' training histories, reading it with label-free indices of frequency, position, form and usage and an injected-error test, all computed from sampled outputs. We find that zero-RL almost never coins the marker: on bases that lack the word, RL raises MATH-500 accuracy substantially while {wait} stays near absent. It arrives with training data: mid-training in OLMo-3, and fine-tuning on reasoning traces installs it in most models whose base lacks it. Where the base already uses it, both released RL-Zero recipes (OLMo-3, MiMo) multiply it, whereas KL-free GRPO from the same base does not; OLMo-3's recipe wears down its form and makes it a more sensitive but more liberal alarm; RL after supervised fine-tuning barely changes it, and our GRPO recipe erases a marker seeded by fine-tuning. Finally, we show that marker frequency alone does not indicate RL-induced reasoning and caution against LLM-as-a-judge revision metrics, as automated judges can reverse human-labelled trends.

Multilingual large language models often struggle to reason in low- to mid-resource languages. Prior work has shown that translation can improve multilingual reasoning by helping models access stronger English-centric representations. This raises a central question: How much translation is needed for multilingual large language models to reason reliably, and when does more translation instead trigger interference and overconfidence? Using LLaMA and Qwen models, we run extensive experiments varying text scope and language mode (target-only, English-only, bilingual) to evaluate both accuracy and reliability. Our results reveal a clear trade-off: English context often improve understanding and recover errors caused by non-English comprehension, yet adding redundant bilingual context intensifies interference. We address this trade-off with Reliability-Aware Adaptive Inference (RAAI), a training-free test-time framework that (i) performs Expected Calibration Error (ECE)-aware routing and prompt fusion, and (ii) uses a mid-layer Risk Index (RI) to gate sequential reasoning, allocating compute only when it is likely to help and suppressing harmful bilingual redundancy. Across two model families, RAAI enhances accuracy by 25-37.7% on low-resource languages and lowers calibration error by approximately 3-6%, with the most pronounced benefits in the lowest-resource language tiers.

We address the problem of anomaly detection, that is, detecting anomalous events in a video sequence. Anomaly detection methods based on convolutional neural networks (CNNs) typically leverage proxy tasks, such as reconstructing input video frames, to learn models describing normality without seeing anomalous samples at training time, and quantify the extent of abnormalities using the reconstruction error at test time. The main drawbacks of these approaches are that they do not consider the diversity of normal patterns explicitly, and the powerful representation capacity of CNNs allows to reconstruct abnormal video frames. To address this problem, we present an unsupervised learning approach to anomaly detection that considers the diversity of normal patterns explicitly, while lessening the representation capacity of CNNs. To this end, we propose to use a memory module with a new update scheme where items in the memory record prototypical patterns of normal data. We also present novel feature compactness and separateness losses to train the memory, boosting the discriminative power of both memory items and deeply learned features from normal data. Experimental results on standard benchmarks demonstrate the effectiveness and efficiency of our approach, which outperforms the state of the art.

We consider the problem of learning a target function corresponding to a single hidden layer neural network, with a quadratic activation function after the first layer, and random weights. We consider the asymptotic limit where the input dimension and the network width are proportionally large. Recent work [Cui & al '23] established that linear regression provides Bayes-optimal test error to learn such a function when the number of available samples is only linear in the dimension. That work stressed the open challenge of theoretically analyzing the optimal test error in the more interesting regime where the number of samples is quadratic in the dimension. In this paper, we solve this challenge for quadratic activations and derive a closed-form expression for the Bayes-optimal test error. We also provide an algorithm, that we call GAMP-RIE, which combines approximate message passing with rotationally invariant matrix denoising, and that asymptotically achieves the optimal performance. Technically, our result is enabled by establishing a link with recent works on optimal denoising of extensive-rank matrices and on the ellipsoid fitting problem. We further show empirically that, in the absence of noise, randomly-initialized gradient descent seems to sample the space of weights, leading to zero training loss, and averaging over initialization leads to a test error equal to the Bayes-optimal one.

We identify label errors in the test sets of 10 of the most commonly-used computer vision, natural language, and audio datasets, and subsequently study the potential for these label errors to affect benchmark results. Errors in test sets are numerous and widespread: we estimate an average of at least 3.3% errors across the 10 datasets, where for example label errors comprise at least 6% of the ImageNet validation set. Putative label errors are identified using confident learning algorithms and then human-validated via crowdsourcing (51% of the algorithmically-flagged candidates are indeed erroneously labeled, on average across the datasets). Traditionally, machine learning practitioners choose which model to deploy based on test accuracy - our findings advise caution here, proposing that judging models over correctly labeled test sets may be more useful, especially for noisy real-world datasets. Surprisingly, we find that lower capacity models may be practically more useful than higher capacity models in real-world datasets with high proportions of erroneously labeled data. For example, on ImageNet with corrected labels: ResNet-18 outperforms ResNet-50 if the prevalence of originally mislabeled test examples increases by just 6%. On CIFAR-10 with corrected labels: VGG-11 outperforms VGG-19 if the prevalence of originally mislabeled test examples increases by just 5%. Test set errors across the 10 datasets can be viewed at https://labelerrors.com and all label errors can be reproduced by https://github.com/cleanlab/label-errors.

Since experiencing domain shifts during test-time is inevitable in practice, test-time adaption (TTA) continues to adapt the model after deployment. Recently, the area of continual and gradual test-time adaptation (TTA) emerged. In contrast to standard TTA, continual TTA considers not only a single domain shift, but a sequence of shifts. Gradual TTA further exploits the property that some shifts evolve gradually over time. Since in both settings long test sequences are present, error accumulation needs to be addressed for methods relying on self-training. In this work, we propose and show that in the setting of TTA, the symmetric cross-entropy is better suited as a consistency loss for mean teachers compared to the commonly used cross-entropy. This is justified by our analysis with respect to the (symmetric) cross-entropy's gradient properties. To pull the test feature space closer to the source domain, where the pre-trained model is well posed, contrastive learning is leveraged. Since applications differ in their requirements, we address several settings, including having source data available and the more challenging source-free setting. We demonstrate the effectiveness of our proposed method 'robust mean teacher' (RMT) on the continual and gradual corruption benchmarks CIFAR10C, CIFAR100C, and Imagenet-C. We further consider ImageNet-R and propose a new continual DomainNet-126 benchmark. State-of-the-art results are achieved on all benchmarks.

We present an alternative method of evaluating Quality Estimation systems, which is based on a linguistically-motivated Test Suite. We create a test-set consisting of 14 linguistic error categories and we gather for each of them a set of samples with both correct and erroneous translations. Then, we measure the performance of 5 Quality Estimation systems by checking their ability to distinguish between the correct and the erroneous translations. The detailed results are much more informative about the ability of each system. The fact that different Quality Estimation systems perform differently at various phenomena confirms the usefulness of the Test Suite.

Software is used in critical applications in our day-to-day life and it is important to ensure its correctness. One popular approach to assess correctness is to evaluate software on tests. If a test fails, it indicates a fault in the software under test; if all tests pass correctly, one may assume that the software is correct. However, the reliability of these results depends on the test suite considered, and there is a risk of false negatives (i.e. software that passes all available tests but contains bugs because some cases are not tested). Therefore, it is important to consider error-inducing test cases when evaluating software. To support data-driven creation of such a test-suite, which is especially of interest for testing software synthesized from large language models, we curate a dataset (Codehacks) of programming problems together with corresponding error-inducing test cases (i.e., "hacks"). This dataset is collected from the wild, in particular, from the Codeforces online judge platform. The dataset comprises 288,617 hacks for 5,578 programming problems, each with a natural language description, as well as the source code for 2,196 submitted solutions to these problems that can be broken with their corresponding hacks. Keywords: competitive programming, language model, dataset

Large language models (LLMs) have been widely applied to assist test generation with the source code under test provided as the context. This paper aims to answer the question: If the source code under test is incorrect, will LLMs be misguided when generating tests? The effectiveness of test cases is measured by their accuracy, coverage, and bug detection effectiveness. Our evaluation results with five open- and six closed-source LLMs on four datasets demonstrate that incorrect code can significantly mislead LLMs in generating correct, high-coverage, and bug-revealing tests. For instance, in the HumanEval dataset, LLMs achieve 80.45% test accuracy when provided with task descriptions and correct code, but only 57.12% when given task descriptions and incorrect code. For the APPS dataset, prompts with correct code yield tests that detect 39.85% of the bugs, while prompts with incorrect code detect only 19.61%. These findings have important implications for the deployment of LLM-based testing: using it on mature code may help protect against future regression, but on early-stage immature code, it may simply bake in errors. Our findings also underscore the need for further research to improve LLMs resilience against incorrect code in generating reliable and bug-revealing tests.

We study a fundamental problem in the evaluation of large language models that we call training on the test task. Unlike wrongful practices like training on the test data, leakage, or data contamination, training on the test task is not a malpractice. Rather, the term describes a growing set of practices that utilize knowledge about evaluation tasks at training time. We demonstrate that training on the test task confounds both relative model evaluations and claims about emergent capabilities. We argue that the seeming superiority of one model family over another may be explained by a different degree of training on the test task. To this end, we propose an effective method to adjust for the effect of training on the test task on benchmark evaluations. Put simply, to fine-tune each model under comparison on the same task-relevant data prior to evaluation. We then show that instances of emergent behavior disappear gradually as models train on the test task. Our work promotes a new perspective on the evaluation of large language models, with broad implications for benchmarking and the study of emergent capabilities.

Recent advances in automated test generation utilises language models to produce unit tests. While effective, language models tend to generate many incorrect tests with respect to both syntax and semantics. Although such incorrect tests can be easily detected and discarded, they constitute a "missed opportunity" -- if fixed, they are often valuable as they directly add testing value (they effectively target the underlying program logic to be tested) and indirectly form good seeds for generating additional tests. To this end, we propose a simple technique for repairing some of these incorrect tests through a combination of rule-based static analysis and re-prompting. We evaluate this simple approach, named YATE, on a set of 6 open-source projects and show that it can effectively produce tests that cover on average 32.06% more lines and kill 21.77% more mutants than a plain LLM-based method. We also compare YATE with four other LLM-based methods, namely HITS, SYMPROMPT, TESTSPARK and COVERUP and show that it produces tests that cover substantially more code. YATE achieves 22% higher line coverage, 20% higher branch coverage and kill 20% more mutants at a comparable cost (number of calls to LLMs).

A large share of day-to-day work in professional domains happens inside PDF files: benefits packets, leases, datasheets, clinical guidelines, construction plans. Benchmarks for document AI have generally measured the required capabilities in isolation: OCR, layout analysis, chart reasoning, table QA, document VQA. A high score on any one of them does not necessarily reveal whether a model can answer a realistic question that someone in the field would actually ask about a specific PDF. GDP_pdf is a benchmark built to measure this directly. It consists of question-document pairs authored by working professionals in ten fields, and a candidate question was kept only when at least two frontier multimodal models failed it in a way that mattered: a wrong answer, missed decisive evidence, or a fabricated claim, rather than a superficial difference such as style. Each item comes with a rubric of atomic criteria, so we can report a graded rubric score as well as a strict task-level pass rate, and each item is tagged against a taxonomy of eleven capabilities in three tiers, spanning text extraction and grounding, table and chart comprehension, cross-referencing, spatial reasoning, and abstention on unsupported queries. We report results for seventeen frontier models on the 100-item benchmark: the best model passes only 30.7% of the items and the worst passes 2%. Most errors trace back to a small set of recurring loss patterns: misaligned tables, misread charts, skipped footnotes and exclusions, miscounted floor-plan symbols, scan noise, and amendments that supersede earlier text.

Language models have demonstrated remarkable performance in solving reasoning tasks; however, even the strongest models still occasionally make reasoning mistakes. Recently, there has been active research aimed at improving reasoning accuracy, particularly by using pretrained language models to "self-correct" their mistakes via multi-round prompting. In this paper, we follow this line of work but focus on understanding the usefulness of incorporating "error-correction" data directly into the pretraining stage. This data consists of erroneous solution steps immediately followed by their corrections. Using a synthetic math dataset, we show promising results: this type of pretrain data can help language models achieve higher reasoning accuracy directly (i.e., through simple auto-regression, without multi-round prompting) compared to pretraining on the same amount of error-free data. We also delve into many details, such as (1) how this approach differs from beam search, (2) how such data can be prepared, (3) whether masking is needed on the erroneous tokens, (4) the amount of error required, (5) whether such data can be deferred to the fine-tuning stage, and many others.

Noise plagues many numerical datasets, where the recorded values in the data may fail to match the true underlying values due to reasons including: erroneous sensors, data entry/processing mistakes, or imperfect human estimates. We consider general regression settings with covariates and a potentially corrupted response whose observed values may contain errors. By accounting for various uncertainties, we introduced veracity scores that distinguish between genuine errors and natural data fluctuations, conditioned on the available covariate information in the dataset. We propose a simple yet efficient filtering procedure for eliminating potential errors, and establish theoretical guarantees for our method. We also contribute a new error detection benchmark involving 5 regression datasets with real-world numerical errors (for which the true values are also known). In this benchmark and additional simulation studies, our method identifies incorrect values with better precision/recall than other approaches.

Large LMs such as GPT-3 are powerful, but can commit mistakes that are obvious to humans. For example, GPT-3 would mistakenly interpret "What word is similar to good?" to mean a homophone, while the user intended a synonym. Our goal is to effectively correct such errors via user interactions with the system but without retraining, which will be prohibitively costly. We pair GPT-3 with a growing memory of recorded cases where the model misunderstood the user's intents, along with user feedback for clarification. Such a memory allows our system to produce enhanced prompts for any new query based on the user feedback for error correction on similar cases in the past. On four tasks (two lexical tasks, two advanced ethical reasoning tasks), we show how a (simulated) user can interactively teach a deployed GPT-3, substantially increasing its accuracy over the queries with different kinds of misunderstandings by the GPT-3. Our approach is a step towards the low-cost utility enhancement for very large pre-trained LMs. Code, data, and instructions to implement MEMPROMPT for a new task at https://www.memprompt.com/.

Large Language Models (LLM) benchmarks tell us when models fail, but not why they fail. A wrong answer on a reasoning dataset may stem from formatting issues, calculation errors, or dataset noise rather than weak reasoning. Without disentangling such causes, benchmarks remain incomplete and cannot reliably guide model improvement. We introduce ErrorMap, the first method to chart the sources of LLM failure. It extracts a model's unique "failure signature", clarifies what benchmarks measure, and broadens error identification to reduce blind spots. This helps developers debug models, aligns benchmark goals with outcomes, and supports informed model selection. ErrorMap works on any model or dataset with the same logic. Applying our method to 35 datasets and 83 models we generate ErrorAtlas, a taxonomy of model errors, revealing recurring failure patterns. ErrorAtlas highlights error types that are currently underexplored in LLM research, such as omissions of required details in the output and question misinterpretation. By shifting focus from where models succeed to why they fail, ErrorMap and ErrorAtlas enable advanced evaluation - one that exposes hidden weaknesses and directs progress. Unlike success, typically measured by task-level metrics, our approach introduces a deeper evaluation layer that can be applied globally across models and tasks, offering richer insights into model behavior and limitations. We make the taxonomy and code publicly available with plans to periodically update ErrorAtlas as new benchmarks and models emerge.

Advances in natural language processing, such as transfer learning from pre-trained language models, have impacted how models are trained for programming language tasks too. Previous research primarily explored code pre-training and expanded it through multi-modality and multi-tasking, yet the data for downstream tasks remain modest in size. Focusing on data utilization for downstream tasks, we propose and adapt augmentation methods that yield consistent improvements in code translation and summarization by up to 6.9% and 7.5% respectively. Further analysis suggests that our methods work orthogonally and show benefits in output code style and numeric consistency. We also discuss test data imperfections.

Large pre-trained language models have brought remarkable progress in NLP. Pre-training and Fine-tuning have given state-of-art performance across tasks in text processing. Data Augmentation techniques have also helped build state-of-art models on low or zero resource tasks. Many works in the past have attempted at learning a single massively-multilingual machine translation model for zero-shot translation. Although those translation models are producing correct translations, the main challenge is those models are producing the wrong languages for zero-shot translation. This work and its results indicate that prompt conditioned large models do not suffer from off-target language errors i.e. errors arising due to translation to wrong languages. We empirically demonstrate the effectiveness of self-supervised pre-training and data augmentation for zero-shot multi-lingual machine translation.

Verifying the final answer of a long reasoning trace may show that the trace is wrong but it does not distinguish three cases. The problem may lie beyond the model's capability, another attempt would succeed, or one step in an otherwise recoverable trace made the trace fail. Standard hidden-state probes are trained on final-answer correctness, a label that conflates these cases. Monte Carlo rollouts distinguish the cases and locate the failing step at additional inference cost. We test the hidden states of 4B and 8B reasoning models for the failure type and the failing step and compare each probe with the strongest baseline that uses no hidden states. On incorrect 8B traces, the confidence probe, a calibrated correctness probe at a midpoint cutoff, misses 81% of cases in which a single step makes the trace unrecoverable and 47% of cases in which another attempt would succeed. A probe with a different training design over the same states reverses the ordering of these two kinds of failure, so either ordering can be read from those states. At both scales, no single tested probe locates the failing step more accurately than a rule based only on trace length, and at 4B a probe trained to predict that step fails to outperform this rule in any of 10 runs on identical traces. The confidence probe therefore indicates that an answer is wrong, but its failure-type ordering depends on its training design.

Recent works have shown the benefits to LLMs from fine-tuning golden-standard Chain-of-Thought (CoT) rationales or using them as correct examples in few-shot prompting. While humans can indeed imitate correct examples, learning from our mistakes is another vital aspect of human cognition. Hence, a question naturally arises: can LLMs learn and benefit from their mistakes, especially for their reasoning? This study investigates this problem from both the prompting and model-tuning perspectives. We begin by introducing CoTErrorSet, a new benchmark with 609,432 questions, each designed with both correct and error references, and demonstrating the types and reasons for making such mistakes. To explore the effectiveness of those mistakes, we design two methods: (1) Self-rethinking prompting guides LLMs to rethink whether they have made similar previous mistakes; and (2) Mistake tuning involves finetuning models in both correct and incorrect reasoning domains, rather than only tuning models to learn ground truth in traditional methodology. We conduct a series of experiments to prove LLMs can obtain benefits from mistakes in both directions. Our two methods offer potentially cost-effective strategies by leveraging errors to enhance reasoning capabilities, which costs significantly less than creating meticulously hand-crafted golden references. We ultimately make a thorough analysis of the reasons behind LLMs' errors, which provides directions that future research needs to overcome. CoTErrorSet will be published soon on \url{https://github.com/YookiTong/Learn-from-Mistakes-CotErrorSet}.

Large language models can diagnose a student program from its code, problem statement, and reference solution. Evaluating this ability requires a clear definition of what counts as the correct diagnosis. We introduce PyMETA, a Python error dataset with 48,646 student submissions to 155 problems. Every submission has a single label for the first execution error identified by an Online Judge, or No Error when the program passes all tests. A targeted subset of 97 submissions also has expert labels collected through iterative repair and re-execution. The taxonomy has three levels; its most detailed level contains 14 labels, including No Error, Logic Error, named Python exceptions, and an Other Errors category. We evaluate two finetuned models and two groups of prompted LLMs: four earlier models and four recent models. When evaluated against the first execution error, the recent prompted models reach 87.5--93.8% macro F1, above the strongest finetuned baseline at 80.6%. This is the opposite of the comparison obtained with the earlier prompted models. On the 97-item expert subset, however, exact-set match is only 43.3--48.5%, although sample F1 is about 79--81%. Output format also matters. On the same 45 audited submissions whose expert label sets do not contain Logic Error, none of the four recent models returns that label under single-error prompting, but 46.7--57.8% of their multi-error outputs include it, usually after an explicit-error label. The results show that model rankings and claims about label bias depend on the meaning of the gold label, the number of labels a model may return, and the scoring rule.

Tests that cause spurious failures without any code changes, i.e., flaky tests, hamper regression testing, increase maintenance costs, may shadow real bugs, and decrease trust in tests. While the prevalence and importance of flakiness is well established, prior research focused on Java projects, thus raising the question of how the findings generalize. In order to provide a better understanding of the role of flakiness in software development beyond Java, we empirically study the prevalence, causes, and degree of flakiness within software written in Python, one of the currently most popular programming languages. For this, we sampled 22352 open source projects from the popular PyPI package index, and analyzed their 876186 test cases for flakiness. Our investigation suggests that flakiness is equally prevalent in Python as it is in Java. The reasons, however, are different: Order dependency is a much more dominant problem in Python, causing 59% of the 7571 flaky tests in our dataset. Another 28% were caused by test infrastructure problems, which represent a previously undocumented cause of flakiness. The remaining 13% can mostly be attributed to the use of network and randomness APIs by the projects, which is indicative of the type of software commonly written in Python. Our data also suggests that finding flaky tests requires more runs than are often done in the literature: A 95% confidence that a passing test case is not flaky on average would require 170 reruns.

Chain-of-thought (CoT) prompting has become central to mathematical reasoning in large language models, yet models remain brittle to early errors: a single arithmetic slip or unjustified inference typically propagates uncorrected to an incorrect final answer. We investigate whether training on intentionally flawed reasoning traces can teach models to detect and recover from such errors without degrading standard problem-solving ability. Using competition-level problems from MATH-lighteval, we generate CoT prefixes containing exactly one controlled error, either a calculation error (sign flips, dropped terms) or a reasoning error (misapplied rules, unjustified logical steps), and fine-tune Qwen3-4B with GRPO using a binary final-answer reward. Our Mixed-CoT-RL model matches standard RL on clean problems (41% vs 41%) while substantially outperforming it on problems prefilled with flawed reasoning (24% vs 19%). Notably, clean-only RL fine-tuning degrades robustness below the untuned baseline 19% vs. 20%), indicating that conventional training increases susceptibility to misleading prefills. Among error types, training on reasoning errors yields greater robustness gains than calculation errors alone, with mixed training performing best. These findings demonstrate that exposure to flawed traces during training can improve error-recovery behavior without sacrificing accuracy, suggesting a path toward more robust mathematical reasoning in LLMs.

Research on reasoning in language models (LMs) predominantly focuses on improving the correctness of their outputs. But some important applications require modeling reasoning patterns that are incorrect. For example, automated systems that can reason about and simulate student errors are useful for providing real-time feedback in the classroom or offline practice for educators-in-training. This paper presents a new method, MISTAKE, that (1) constructs high-quality synthetic examples of reasoning errors by leveraging cycle consistency between incorrect answers and latent misconceptions; and (2) uses the generated data to learn models for student simulation, misconception classification, and answer generation. We evaluate MISTAKE on three educational tasks and find that it results in (1) higher accuracy when simulating incorrect student answers based on specific misconceptions, (2) increased performance inferring latent misconceptions from observed incorrect answers, and (3) higher alignment with expert-written distractor answers when generating incorrect answers (e.g., for multiple-choice tests).

The MEDIQA-CORR 2024 shared task aims to assess the ability of Large Language Models (LLMs) to identify and correct medical errors in clinical notes. In this study, we evaluate the capability of general LLMs, specifically GPT-3.5 and GPT-4, to identify and correct medical errors with multiple prompting strategies. Recognising the limitation of LLMs in generating accurate corrections only via prompting strategies, we propose incorporating error-span predictions from a smaller, fine-tuned model in two ways: 1) by presenting it as a hint in the prompt and 2) by framing it as multiple-choice questions from which the LLM can choose the best correction. We found that our proposed prompting strategies significantly improve the LLM's ability to generate corrections. Our best-performing solution with 8-shot + CoT + hints ranked sixth in the shared task leaderboard. Additionally, our comprehensive analyses show the impact of the location of the error sentence, the prompted role, and the position of the multiple-choice option on the accuracy of the LLM. This prompts further questions about the readiness of LLM to be implemented in real-world clinical settings.

In industrial settings, surface defects on steel can significantly compromise its service life and elevate potential safety risks. Traditional defect detection methods predominantly rely on manual inspection, which suffers from low efficiency and high costs. Although automated defect detection approaches based on Convolutional Neural Networks(e.g., Mask R-CNN) have advanced rapidly, their reliability remains challenged due to data annotation uncertainties during deep model training and overfitting issues. These limitations may lead to detection deviations when processing the given new test samples, rendering automated detection processes unreliable. To address this challenge, we first evaluate the detection model's practical performance through calibration data that satisfies the independent and identically distributed (i.i.d) condition with test data. Specifically, we define a loss function for each calibration sample to quantify detection error rates, such as the complement of recall rate and false discovery rate. Subsequently, we derive a statistically rigorous threshold based on a user-defined risk level to identify high-probability defective pixels in test images, thereby constructing prediction sets (e.g., defect regions). This methodology ensures that the expected error rate (mean error rate) on the test set remains strictly bounced by the predefined risk level. Additionally, we observe a negative correlation between the average prediction set size and the risk level on the test set, establishing a statistically rigorous metric for assessing detection model uncertainty. Furthermore, our study demonstrates robust and efficient control over the expected test set error rate across varying calibration-to-test partitioning ratios, validating the method's adaptability and operational effectiveness.

High-quality, error-free datasets are a key ingredient in building reliable, accurate, and unbiased machine learning (ML) models. However, real world datasets often suffer from errors due to sensor malfunctions, data entry mistakes, or improper data integration across multiple sources that can severely degrade model performance. Detecting and correcting these issues typically require tailor-made solutions and demand extensive domain expertise. Consequently, automation is challenging, rendering the process labor-intensive and tedious. In this study, we investigate whether Large Language Models (LLMs) can help alleviate the burden of manual data cleaning. We set up an experiment in which an LLM, paired with Python, is tasked with cleaning the training dataset to improve the performance of a learning algorithm without having the ability to modify the training pipeline or perform any feature engineering. We run this experiment on multiple Kaggle datasets that have been intentionally corrupted with errors. Our results show that LLMs can identify and correct erroneous entries, such as illogical values or outlier, by leveraging contextual information from other features within the same row, as well as feedback from previous iterations. However, they struggle to detect more complex errors that require understanding data distribution across multiple rows, such as trends and biases.

Flaky tests are software tests that exhibit a seemingly random outcome (pass or fail) when run against the same, identical code. Previous work has examined fixes to flaky tests and has proposed automated solutions to locate as well as fix flaky tests--we complement it by examining the perceptions of software developers about the nature, relevance, and challenges of this phenomenon. We asked 21 professional developers to classify 200 flaky tests they previously fixed, in terms of the nature of the flakiness, the origin of the flakiness, and the fixing effort. We complement this analysis with information about the fixing strategy. Subsequently, we conducted an online survey with 121 developers with a median industrial programming experience of five years. Our research shows that: The flakiness is due to several different causes, four of which have never been reported before, despite being the most costly to fix; flakiness is perceived as significant by the vast majority of developers, regardless of their team's size and project's domain, and it can have effects on resource allocation, scheduling, and the perceived reliability of the test suite; and the challenges developers report to face regard mostly the reproduction of the flaky behavior and the identification of the cause for the flakiness. Data and materials [https://doi.org/10.5281/zenodo.3265785].

AI agents can externalize what they learn from past tasks into reusable skills, such as procedures, checklists, code, or other executable artifacts, that can be retrieved and reused when solving new tasks. Self-evolving skill methods keep rewriting these skills after each round of practice on training tasks, and the skill is then used on new tasks of the same kind. We ask a question: does the improvement a skill shows on its training tasks carry over to new test tasks? We test five self-evolving methods and a one-shot skill on six benchmarks, with the same model, the same agent, and the same train/test split for every method. Of the 21 skills that improve on their training tasks, 5 keep all of that improvement on the test tasks, 13 keep part of it, and 3 keep none of it. No existing method is best everywhere. When we read the skills, the ones that carry over badly often fix details that should depend on the task, such as column names and output files, or turn a fix for one failure into a rule for every task. An LLM judge that reads the skill content can often see this: it ranks finished skills the same way the test results do in 86\% of pairs. But it predicts the effect of a single edit poorly, so edits still have to be tested by running them. Based on these findings, we describe Generalizable Skill Optimization (GSO), which keeps only a guide for writing skills and writes a new skill for each task; it scores highest on all six benchmarks.

In software testing, a set of test cases is constructed according to some predefined selection criteria. The software is then examined against these test cases. Three interesting observations have been made on the current artifacts of software testing. Firstly, an error-revealing test case is considered useful while a successful test case which does not reveal software errors is usually not further investigated. Whether these successful test cases still contain useful information for revealing software errors has not been properly studied. Secondly, no matter how extensive the testing has been conducted in the development phase, errors may still exist in the software [5]. These errors, if left undetected, may eventually cause damage to the production system. The study of techniques for uncovering software errors in the production phase is seldom addressed in the literature. Thirdly, as indicated by Weyuker in [6], the availability of test oracles is pragmatically unattainable in most situations. However, the availability of test oracles is generally assumed in conventional software testing techniques. In this paper, we propose a novel test case selection technique that derives new test cases from the successful ones. The selection aims at revealing software errors that are possibly left undetected in successful test cases which may be generated using some existing strategies. As such, the proposed technique augments the effectiveness of existing test selection strategies. The technique also helps uncover software errors in the production phase and can be used in the absence of test oracles.

Widely used learned metrics for machine translation evaluation, such as COMET and BLEURT, estimate the quality of a translation hypothesis by providing a single sentence-level score. As such, they offer little insight into translation errors (e.g., what are the errors and what is their severity). On the other hand, generative large language models (LLMs) are amplifying the adoption of more granular strategies to evaluation, attempting to detail and categorize translation errors. In this work, we introduce xCOMET, an open-source learned metric designed to bridge the gap between these approaches. xCOMET integrates both sentence-level evaluation and error span detection capabilities, exhibiting state-of-the-art performance across all types of evaluation (sentence-level, system-level, and error span detection). Moreover, it does so while highlighting and categorizing error spans, thus enriching the quality assessment. We also provide a robustness analysis with stress tests, and show that xCOMET is largely capable of identifying localized critical errors and hallucinations.

We introduce a few-shot learning framework for error detection. We show that data augmentation (a form of weak supervision) is key to training high-quality, ML-based error detection models that require minimal human involvement. Our framework consists of two parts: (1) an expressive model to learn rich representations that capture the inherent syntactic and semantic heterogeneity of errors; and (2) a data augmentation model that, given a small seed of clean records, uses dataset-specific transformations to automatically generate additional training data. Our key insight is to learn data augmentation policies from the noisy input dataset in a weakly supervised manner. We show that our framework detects errors with an average precision of ~94% and an average recall of ~93% across a diverse array of datasets that exhibit different types and amounts of errors. We compare our approach to a comprehensive collection of error detection methods, ranging from traditional rule-based methods to ensemble-based and active learning approaches. We show that data augmentation yields an average improvement of 20 F1 points while it requires access to 3x fewer labeled examples compared to other ML approaches.

Large language models (LLMs) have demonstrated remarkable reasoning capability in solving mathematical problems. However, existing approaches primarily focus on improving the quality of correct training data, e.g., distilling high-quality correct solutions from advanced models, neglecting the value contained in error data, potentially hindering the model's reflective ability. Though some studies attempt to leverage error data, they often involve complex mechanisms, such as Monte Carlo Tree Search (MCTS) to explore error nodes. In this work, we propose to enhance LLMs' reasoning ability by Learning from Errors for Mathematical Advancement (LEMMA). LEMMA constructs data consisting of an incorrect solution with an erroneous step and a reflection connection to a correct solution for fine-tuning. Specifically, we systematically analyze the model-generated error types and introduce an error-type grounded mistake augmentation method to collect diverse and representative errors. Correct solutions are either from fixing the errors or generating a fresh start. Through a model-aware smooth reflection connection, the erroneous solution is transferred to the correct one. By fine-tuning on the constructed dataset, the model is able to self-correct errors autonomously within the generation process without relying on external critique models. Experimental results demonstrate that LEMMA achieves significant performance improvements over other strong baselines.

Reasoning models (e.g., DeepSeek-R1) generate long chains of thought to solve harder problems, but they often loop, repeating the same text at low temperatures or with greedy decoding. We study why this happens and what role temperature plays. With open reasoning models, we find that looping is common at low temperature. Larger models tend to loop less, and distilled students loop significantly even when their teachers rarely do. This points to mismatches between the training distribution and the learned model, which we refer to as errors in learning, as a key cause. To understand how such errors cause loops, we introduce a synthetic graph reasoning task and demonstrate two mechanisms. First, risk aversion caused by hardness of learning: when the correct progress-making action is hard to learn but an easy cyclic action is available, the model puts relatively more probability on the cyclic action and gets stuck. Second, even when there is no hardness, Transformers show an inductive bias toward temporally correlated errors, so the same few actions keep being chosen and loops appear. Higher temperature reduces looping by promoting exploration, but it does not fix the errors in learning, so generations remain much longer than necessary at high temperature; in this sense, temperature is a stopgap rather than a holistic solution. We end with a discussion of training-time interventions aimed at directly reducing errors in learning.

With Large Language Models (LLMs) being widely used across various tasks, detecting errors in their responses is increasingly crucial. However, little research has been conducted on error detection of LLM responses. Collecting error annotations on LLM responses is challenging due to the subjective nature of many NLP tasks, and thus previous research focuses on tasks of little practical value (e.g., word sorting) or limited error types (e.g., faithfulness in summarization). This work introduces ReaLMistake, the first error detection benchmark consisting of objective, realistic, and diverse errors made by LLMs. ReaLMistake contains three challenging and meaningful tasks that introduce objectively assessable errors in four categories (reasoning correctness, instruction-following, context-faithfulness, and parameterized knowledge), eliciting naturally observed and diverse errors in responses of GPT-4 and Llama 2 70B annotated by experts. We use ReaLMistake to evaluate error detectors based on 12 LLMs. Our findings show: 1) Top LLMs like GPT-4 and Claude 3 detect errors made by LLMs at very low recall, and all LLM-based error detectors perform much worse than humans. 2) Explanations by LLM-based error detectors lack reliability. 3) LLMs-based error detection is sensitive to small changes in prompts but remains challenging to improve. 4) Popular approaches to improving LLMs, including self-consistency and majority vote, do not improve the error detection performance. Our benchmark and code are provided at https://github.com/psunlpgroup/ReaLMistake.

This paper proposes one of the first clinical applications of multimodal large language models (LLMs) as an assistant for radiologists to check errors in their reports. We created an evaluation dataset from two real-world radiology datasets (MIMIC-CXR and IU-Xray), with 1,000 subsampled reports each. A subset of original reports was modified to contain synthetic errors by introducing various type of mistakes. The evaluation contained two difficulty levels: SIMPLE for binary error-checking and COMPLEX for identifying error types. LLaVA (Large Language and Visual Assistant) variant models, including our instruction-tuned model, were used for the evaluation. Additionally, a domain expert evaluation was conducted on a small test set. At the SIMPLE level, the LLaVA v1.5 model outperformed other publicly available models. Instruction tuning significantly enhanced performance by 47.4% and 25.4% on MIMIC-CXR and IU-Xray data, respectively. The model also surpassed the domain experts accuracy in the MIMIC-CXR dataset by 1.67%. Notably, among the subsets (N=21) of the test set where a clinician did not achieve the correct conclusion, the LLaVA ensemble mode correctly identified 71.4% of these cases. This study marks a promising step toward utilizing multi-modal LLMs to enhance diagnostic accuracy in radiology. The ensemble model demonstrated comparable performance to clinicians, even capturing errors overlooked by humans. Nevertheless, future work is needed to improve the model ability to identify the types of inconsistency.

Consistently checking the statistical significance of experimental results is one of the mandatory methodological steps to address the so-called "reproducibility crisis" in deep reinforcement learning. In this tutorial paper, we explain how the number of random seeds relates to the probabilities of statistical errors. For both the t-test and the bootstrap confidence interval test, we recall theoretical guidelines to determine the number of random seeds one should use to provide a statistically significant comparison of the performance of two algorithms. Finally, we discuss the influence of deviations from the assumptions usually made by statistical tests. We show that they can lead to inaccurate evaluations of statistical errors and provide guidelines to counter these negative effects. We make our code available to perform the tests.

An unsuccessful LLM agent rollout contains more information than its final reward: the observations available to the agent, the actions it chose, and the environment's responses. Reusing this experience for learning requires identifying a decision to revise and testing a concrete alternative. We introduce the Agent Error Dataset (AED), comprising 50,228 error-diagnosis pairs from 9,961 source tasks across 33 environments, 19 harness families, and 23 policy models in text-based agent systems. We retain source traces and execution metadata to support cross-setting failure analysis and re-diagnosis without repeating the original rollout. Our five-stage Agentic Error-to-Training (AET) pipeline collects natural failures, generates diagnoses and proposed corrections, and checks them against recorded evidence. Where replay is supported, we compare corrections with original-action retries from the same checkpoint under matched execution settings. We then construct separate training views for diagnosis and actor recovery. Across 3,062 matched replay pairs, first-proposal corrections raise verifier pass rates from 18.4% to 51.1%, a gain of 32.7 percentage points. Using a separately frozen diagnosis release, full-diagnosis fine-tuning on 1,656 source tasks raises Qwen3-8B's exact-step agreement with internal teacher labels from 47.2% to 63.6%, averaged over three seeds on a 943-case holdout. The strongest prompted reference in this comparison scores 54.7%, and mean agreement improves at each of four increasing training-set sizes. In a single-seed comparison of actor-training recipes, action-only repair training scores 6.67 percentage points higher on WebShop-lite than success-only training.

An agent harness, the code that turns a model into an agent, writes its own record of each run, and that record is all a later reader gets when a run is disputed, investigated or audited. We call a record evidentiary when a reader who was not there can check it without trusting the writer. Across sixteen deployed frameworks, none writes one in full. Hearsay examines the record, not the task: five harnesses run fourteen tasks, three blinded LLM examiners and a human panel read the records, and every excerpt an examiner quotes is checked mechanically for who wrote it. First, the record lets a reader name the fault but not prove how the run went. Examiners name the right fault in 74 to 91% of 140 runs, but the fault can be proved only from two files the benchmark adds; for what happened in between, fewer than one citation in ten lands on anything the harness did not write, and the examiner with the fewest false alarms catches half of the entries we delete, rewrite or fabricate. Second, the remedy is a second author, not a stronger seal on the first. An append-only log of what passes between harness and model, kept outside the harness and read against the record in both directions, reports all 28 omissions and fabrications we made a harness commit as it ran, where a hash chain over the harness's own record passes all 28. Handed the log, examiners keep their fault verdicts but rest more of their citations on what the harness did not write. What makes a record evidence is who writes it, not what is captured.

Automatic Question Generation (QG) often produces outputs with critical defects, such as factual hallucinations and answer mismatches. However, existing evaluation methods, including LLM-based evaluators, mainly adopt a black-box and holistic paradigm without explicit error modeling, leading to the neglect of such defects and overestimation of question quality. To address this issue, we propose ErrEval, a flexible and Error-aware Evaluation framework that enhances QG evaluation through explicit error diagnostics. Specifically, ErrEval reformulates evaluation as a two-stage process of error diagnosis followed by informed scoring. At the first stage, a lightweight plug-and-play Error Identifier detects and categorizes common errors across structural, linguistic, and content-related aspects. These diagnostic signals are then incorporated as explicit evidence to guide LLM evaluators toward more fine-grained and grounded judgments. Extensive experiments on three benchmarks demonstrate the effectiveness of ErrEval, showing that incorporating explicit diagnostics improves alignment with human judgments. Further analyses confirm that ErrEval effectively mitigates the overestimation of low-quality questions.

Operator-learning benchmarks ship as full-precision arrays, and they have grown to terabyte scale. A curator who wants to distribute one has to decide how coarsely to store it. That decision is usually made by fixing a tolerance on the reconstruction error of the stored field. We show that this quantity is measured in the wrong place. It compares the stored field with the original, before any model is trained. What the curator is buying is the accuracy of a model trained on the compressed copy, and the two can disagree: two PDEBench families stored to identical field error differ threefold downstream. A solution operator smooths, so only part of the codec error ever reaches the model's output. That part can be measured. Push a compressed field through a model already trained at full precision, and compare its output against the output on the clean field. The measurement costs forward passes and no training on compressed data. It spans more than two orders of magnitude across six PDE families, and it orders downstream cost where field error does not, inverting 12 of 104 cross-dataset comparisons against field error's 36. Read as a budget, the same curve nominates a storage rate for each dataset. A back-check against trained models finds those rates wrong in both directions by up to a factor of two in stored bits.

Although large language models (LLMs) have become transformative, they still make mistakes and can explore unproductive reasoning paths. Self-correction is an important capability for a trustworthy LLM, particularly an autoregressive LLM. While LLMs can identify error in user input, they exhibit a systematic 'Self-Correction Blind Spot' - failing to correct identical error in their own outputs. To systematically study this phenomenon, we introduce Self-Correction Bench, a systematic framework to measure this phenomenon through controlled error injection at three complexity levels. Testing 14 models, we find an average 64.5% blind spot rate. We find multiple evidences that this limitation relates to training data composition: human training demonstrations predominantly show error-free responses rather than error-correction sequences, unlike RL-trained models that learn error correction through outcome feedback. Remarkably, simply appending "Wait" reduces blind spots by 89.3%, suggesting that the capability exists but requires activation. Our work highlights a critical limitation in current LLMs and offers potential avenues for improving their reliability and trustworthiness.

We build new test sets for the CIFAR-10 and ImageNet datasets. Both benchmarks have been the focus of intense research for almost a decade, raising the danger of overfitting to excessively re-used test sets. By closely following the original dataset creation processes, we test to what extent current classification models generalize to new data. We evaluate a broad range of models and find accuracy drops of 3% - 15% on CIFAR-10 and 11% - 14% on ImageNet. However, accuracy gains on the original test sets translate to larger gains on the new test sets. Our results suggest that the accuracy drops are not caused by adaptivity, but by the models' inability to generalize to slightly "harder" images than those found in the original test sets.

We introduce TestCase-Eval, a new benchmark for systematic evaluation of LLMs in test-case generation. TestCase-Eval includes 500 algorithm problems and 100,000 human-crafted solutions from the Codeforces platform. It focuses on two pivotal tasks: (1) Fault Coverage, which measures how well LLM-generated test sets probe diverse input scenarios and cover a wide range of potential failure modes. (2) Fault Exposure, which evaluates whether LLMs can craft a tailored test input that reveals a specific incorrect code implementation. We provide a comprehensive assessment of 19 state-of-the-art open-source and proprietary LLMs on TestCase-Eval, offering insights into their strengths and limitations in generating effective test cases for algorithm problems.

We describe our system for LitTraceQA (GroundLM @ EMNLP 2026): given a research question, retrieve the relevant papers from a pool of 27,487, cite the page and the table or figure where the answer lives, and answer in a requested format. Our main finding is that evidence grounding is limited by retrieval, not by reading. The page selector put the annotator's page, which we call the gold page, in front of the model that locates evidence only about half the time (52.6% gold-page recall), while that model, given the page, cited the right one in 45 of the 48 locators it emitted (94%). When the page was missing it rarely said so: of 45 such cases it returned nothing 14 times, a wrong page 24 times, and a correct page 7 times, so the pipeline failed quietly almost twice as often as it failed visibly. Since the failure was that the right page was never shown, the fix is to stop choosing: each retrieved paper fits in the model's context, so we show it whole. Page ranking survives only as a fallback inside papers too long to fit, which no test-split paper was, and gold-page recall reaches 100% on the papers we can parse. Separately, questions that identify their target by position rather than content, such as "the first author of the 24th reference", are served by parsing rather than retrieval: we resolve the bibliography into an addressable list, which also supplies identifiers the evidence metric scores. The final system scores 0.762 paper F_1, 0.441 evidence F_1 and 0.920 multiple-choice accuracy on the held-out test split. Because the pipeline depends on a closed model without seed control, we release a harness that verifies the paper's central claims against committed artifacts.

A robot that fails at a task faces the first decision in corrective dialogue: act on its own diagnosis, consult another onboard sensor, or interrupt a person. Choosing well requires knowing how much the robot's sensors reveal about the cause and how reliable the robot's own diagnosis is. We build a simulated benchmark in which every failure's true cause is known, because we injected it, and measure what each sensor reveals, with explicit checks against data leakage. Some failures are diagnosable from camera images; others only from the robot's force data (0.99 from force data, no image method above 0.55). We then test six open vision-language models. Their behavior tracks the surface of the prompt, not the evidence: moving the refusal option from last to first in the answer list collapses refusal rates from 78-100% to 0-6% in three of the six swept model-and-family pairs. Accuracy from frames stays at or below a majority-class baseline under every prompt variant, with or without worked examples, and stated confidence carries no information about correctness. Handing the same models the force data as ten lines of text produces the first above-baseline diagnoses, in four of the six models: much of the failure reflects missing sensor data, not missing ability. We pose the choice as a three-action decision problem, act, consult your own sensors, or ask a human, whose optimal policy follows from measured accuracy. The models do not follow it, and their ask rates ignore a fourfold change in question cost. One question to a human still lifts them from that baseline to roughly the answerer's own reliability (0.70-0.81 when they ask). The decision to ask should be tied to measured accuracy and stated costs, not to the model's confidence.

Take three frontier mixture-of-experts models (Alibaba, OpenAI, NVIDIA; 3.6-4.0B active parameters each) and fine-tune them to reason in a low-resource language. On accuracy benchmarks almost nothing happens, and the benchmark itself is noise at this scale: changing only the random seed moves the score by 7.7 points, more than every data and recipe effect we measured. That null is our first result. The real changes live where accuracy cannot see. Base models never think in Greek: 0 of 1,000 reasoning traces, even when the question is Greek, so the model answers correctly while reasoning in a form its user cannot read, audit, or correct. After supervised fine-tuning (SFT), every released checkpoint reasons in the language of the question on ~98% of items, one family at 3x fewer tokens, with judged grammaticality improving on all four models and general ability within a few points of each base: nothing was forgotten, and fluency was gained. We propose six behavioural dimensions that make such changes measurable, each gated to reject any metric that correlates with output length, and we report how our own instruments lied: six failures, each caught by a control. What SFT cannot do is fix its own defects: a quarter of answers skip the requested format, answers leak into the reasoning channel, and an explicit "think in English" is obeyed under half the time. Reinforcement learning with verifiable rewards, pre-registered before training, fixes the first two outright (fallback 24% to 2.5%, leak 3.5% to 0.0%, both against a flat random-reward control) and moves the third (+9.1pp), while the Greek reasoning habit survives an accuracy-only gradient untouched. We release five checkpoints. The instruments, the controls and the pre-registration travel to any low-resource language; Greek is the case that let us measure them.

Ask a language model the same question about the same document twenty times, and it sometimes returns two different answers. We built Probity, a benchmark of 60 tasks and 470 items from real venture-financing filings, to measure how often this happens. Then we audited our own corpus and found a defect any excerpt-built benchmark can carry: items whose evidence is missing from the window of text the model is shown. The audit flags 36 items and separates two failures a single flag would conflate: evidence genuinely absent from the window and answers that must be computed from numbers the window does supply. Flagged items change their answers far more often, wobbling at 0.255 against 0.087 on the 427 clean items, and excluding them cuts apparent cross-model agreement by about a fifth. Before testing whether the missing evidence explains the instability, we registered a prediction: re-cut each window to hold its evidence, and instability should fall below a set threshold. It failed: the repair moved wobble by 0.058, with an interval containing zero. We report the association as correlational. Almost all measurements sit where instability cannot show, which bounds what a corpus built for accuracy can say about stability. We release the corpus, all 112,800 raw responses, and the audit as a runnable check for any document benchmark.

Context: The SZZ algorithm is the de facto standard for labeling bug fixing commits and finding inducing changes for defect prediction data. Recent research uncovered potential problems in different parts of the SZZ algorithm. Most defect prediction data sets provide only static code metrics as features, while research indicates that other features are also important. Objective: We provide an empirical analysis of the defect labels created with the SZZ algorithm and the impact of commonly used features on results. Method: We used a combination of manual validation and adopted or improved heuristics for the collection of defect data. We conducted an empirical study on 398 releases of 38 Apache projects. Results: We found that only half of the bug fixing commits determined by SZZ are actually bug fixing. If a six-month time frame is used in combination with SZZ to determine which bugs affect a release, one file is incorrectly labeled as defective for every file that is correctly labeled as defective. In addition, two defective files are missed. We also explored the impact of the relatively small set of features that are available in most defect prediction data sets, as there are multiple publications that indicate that, e.g., churn related features are important for defect prediction. We found that the difference of using more features is not significant. Conclusion: Problems with inaccurate defect labels are a severe threat to the validity of the state of the art of defect prediction. Small feature sets seem to be a less severe threat.

Large language models (LLMs) possess impressive linguistic capabilities but often fail to faithfully retain factual knowledge, leading to hallucinations and unreliable outputs. Understanding LLMs' knowledge deficiencies by exhaustively evaluating against full-scale knowledge bases is computationally prohibitive, especially for closed-weight models. We propose stochastic error ascent (SEA), a scalable and efficient framework for discovering knowledge deficiencies (errors) in closed-weight LLMs under a strict query budget. Rather than naively probing all knowledge candidates, SEA formulates error discovery as a stochastic optimization process: it iteratively retrieves new high-error candidates by leveraging the semantic similarity to previously observed failures. To further enhance search efficiency and coverage, SEA employs hierarchical retrieval across document and paragraph levels, and constructs a relation directed acyclic graph to model error propagation and identify systematic failure modes. Empirically, SEA uncovers 40.7x more knowledge errors than Automated Capability Discovery and 26.7% more than AutoBencher, while reducing the cost-per-error by 599x and 9x, respectively. Human evaluation confirms the high quality of generated questions, while ablation and convergence analyses validate the contribution of each component in SEA. Further analysis on the discovered errors reveals correlated failure patterns across LLM families and recurring deficits, highlighting the need for better data coverage and targeted fine-tuning in future LLM development.

Several studies showed that Large Language Models (LLMs) can answer medical questions correctly, even outperforming the average human score in some medical exams. However, to our knowledge, no study has been conducted to assess the ability of language models to validate existing or generated medical text for correctness and consistency. In this paper, we introduce MEDEC (https://github.com/abachaa/MEDEC), the first publicly available benchmark for medical error detection and correction in clinical notes, covering five types of errors (Diagnosis, Management, Treatment, Pharmacotherapy, and Causal Organism). MEDEC consists of 3,848 clinical texts, including 488 clinical notes from three US hospital systems that were not previously seen by any LLM. The dataset has been used for the MEDIQA-CORR shared task to evaluate seventeen participating systems [Ben Abacha et al., 2024]. In this paper, we describe the data creation methods and we evaluate recent LLMs (e.g., o1-preview, GPT-4, Claude 3.5 Sonnet, and Gemini 2.0 Flash) for the tasks of detecting and correcting medical errors requiring both medical knowledge and reasoning capabilities. We also conducted a comparative study where two medical doctors performed the same task on the MEDEC test set. The results showed that MEDEC is a sufficiently challenging benchmark to assess the ability of models to validate existing or generated notes and to correct medical errors. We also found that although recent LLMs have a good performance in error detection and correction, they are still outperformed by medical doctors in these tasks. We discuss the potential factors behind this gap, the insights from our experiments, the limitations of current evaluation metrics, and share potential pointers for future research.

Generalizing deep learning models to unknown target domain distribution with low latency has motivated research into test-time training/adaptation (TTT/TTA). Existing approaches often focus on improving test-time training performance under well-curated target domain data. As figured out in this work, many state-of-the-art methods fail to maintain the performance when the target domain is contaminated with strong out-of-distribution (OOD) data, a.k.a. open-world test-time training (OWTTT). The failure is mainly due to the inability to distinguish strong OOD samples from regular weak OOD samples. To improve the robustness of OWTTT we first develop an adaptive strong OOD pruning which improves the efficacy of the self-training TTT method. We further propose a way to dynamically expand the prototypes to represent strong OOD samples for an improved weak/strong OOD data separation. Finally, we regularize self-training with distribution alignment and the combination yields the state-of-the-art performance on 5 OWTTT benchmarks. The code is available at https://github.com/Yushu-Li/OWTTT.

Reproducing a machine learning paper involves most research steps, from installing software and debugging to running experiments, work that AI agents increasingly do. We introduce RECLAIM, a benchmark of 100 NeurIPS 2025 papers that can be rebuilt yearly from new conferences. For each paper we fix in advance the result to reproduce, what counts as a successful reproduction, and a GPU-hour budget. An agent must reproduce that result using the paper and whatever its authors released. What the authors released decides the difficulty tier. Run-tier releases include code, data, and weights; Retrain-tier releases lack weights, so the agent trains the model; Reimplement-tier releases lack code, so the agent writes it. A separate language model grades runs from logs and outputs rather than agents' reports. We run four agents once per paper; the best agent in each tier reproduces only 41% of Run-tier papers, 27% at Retrain, and 15% at Reimplement, where every agent does worst. Failed attempts use on average 29% of their budget, so most stop with budget left. The most common agent error is writing the method without checking any part against the paper's numbers, in 63 of 400 runs.

The paper announces the new long-term challenge for improving the performance of automatic speech recognition systems. The goal of the challenge is to investigate methods of correcting the recognition results on the basis of previously made errors by the speech processing system. The dataset prepared for the task is described and evaluation criteria are presented.

Machine learning methods strive to acquire a robust model during the training process that can effectively generalize to test samples, even in the presence of distribution shifts. However, these methods often suffer from performance degradation due to unknown test distributions. Test-time adaptation (TTA), an emerging paradigm, has the potential to adapt a pre-trained model to unlabeled data during testing, before making predictions. Recent progress in this paradigm has highlighted the significant benefits of using unlabeled data to train self-adapted models prior to inference. In this survey, we categorize TTA into several distinct groups based on the form of test data, namely, test-time domain adaptation, test-time batch adaptation, and online test-time adaptation. For each category, we provide a comprehensive taxonomy of advanced algorithms and discuss various learning scenarios. Furthermore, we analyze relevant applications of TTA and discuss open challenges and promising areas for future research. For a comprehensive list of TTA methods, kindly refer to https://github.com/tim-learn/awesome-test-time-adaptation.

Maybe not. We identify and analyse errors in the popular Massive Multitask Language Understanding (MMLU) benchmark. Even though MMLU is widely adopted, our analysis demonstrates numerous ground truth errors that obscure the true capabilities of LLMs. For example, we find that 57% of the analysed questions in the Virology subset contain errors. To address this issue, we introduce a comprehensive framework for identifying dataset errors using a novel error taxonomy. Then, we create MMLU-Redux, which is a subset of 3,000 manually re-annotated questions across 30 MMLU subjects. Using MMLU-Redux, we demonstrate significant discrepancies with the model performance metrics that were originally reported. Our results strongly advocate for revising MMLU's error-ridden questions to enhance its future utility and reliability as a benchmark. Therefore, we open up MMLU-Redux for additional annotation https://huggingface.co/datasets/edinburgh-dawg/mmlu-redux.

Given the rapid ascent of large language models (LLMs), we study the question: (How) can large language models help in reviewing of scientific papers or proposals? We first conduct some pilot studies where we find that (i) GPT-4 outperforms other LLMs (Bard, Vicuna, Koala, Alpaca, LLaMa, Dolly, OpenAssistant, StableLM), and (ii) prompting with a specific question (e.g., to identify errors) outperforms prompting to simply write a review. With these insights, we study the use of LLMs (specifically, GPT-4) for three tasks: 1. Identifying errors: We construct 13 short computer science papers each with a deliberately inserted error, and ask the LLM to check for the correctness of these papers. We observe that the LLM finds errors in 7 of them, spanning both mathematical and conceptual errors. 2. Verifying checklists: We task the LLM to verify 16 closed-ended checklist questions in the respective sections of 15 NeurIPS 2022 papers. We find that across 119 {checklist question, paper} pairs, the LLM had an 86.6% accuracy. 3. Choosing the "better" paper: We generate 10 pairs of abstracts, deliberately designing each pair in such a way that one abstract was clearly superior than the other. The LLM, however, struggled to discern these relatively straightforward distinctions accurately, committing errors in its evaluations for 6 out of the 10 pairs. Based on these experiments, we think that LLMs have a promising use as reviewing assistants for specific reviewing tasks, but not (yet) for complete evaluations of papers or proposals.

Popular benchmarks (e.g., XNLI) used to evaluate cross-lingual language understanding consist of parallel versions of English evaluation sets in multiple target languages created with the help of professional translators. When creating such parallel data, it is critical to ensure high-quality translations for all target languages for an accurate characterization of cross-lingual transfer. In this work, we find that translation inconsistencies do exist and interestingly they disproportionally impact low-resource languages in XNLI. To identify such inconsistencies, we propose measuring the gap in performance between zero-shot evaluations on the human-translated and machine-translated target text across multiple target languages; relatively large gaps are indicative of translation errors. We also corroborate that translation errors exist for two target languages, namely Hindi and Urdu, by doing a manual reannotation of human-translated test instances in these two languages and finding poor agreement with the original English labels these instances were supposed to inherit.

ASR models often suffer from a long-form deletion problem where the model predicts sequential blanks instead of words when transcribing a lengthy audio (in the order of minutes or hours). From the perspective of a user or downstream system consuming the ASR results, this behavior can be perceived as the model "being stuck", and potentially make the product hard to use. One of the culprits for long-form deletion is training-test data mismatch, which can happen even when the model is trained on diverse and large-scale data collected from multiple application domains. In this work, we introduce a novel technique to simultaneously model different groups of speakers in the audio along with the standard transcript tokens. Speakers are grouped as primary and non-primary, which connects the application domains and significantly alleviates the long-form deletion problem. This improved model neither needs any additional training data nor incurs additional training or inference cost.

A benchmark is not merely a collection of test questions; it naturally induces an empirical measure over difficulty. In this work, we develop this measure-based view of benchmark evaluation. This perspective allows us to identify and formalize two central failure modes: measure mismatch, where the benchmark difficulty distribution differs from the intended difficulty distribution, and coverage gaps, where uncovered regions along the difficulty axis prevent benchmarks from detecting meaningful localized training progress. We propose two diagnostic metrics that capture these two failure modes and provide rigorous theoretical guarantees. Empirically, we analyze 25 popular benchmarks spanning 7 domains using carefully estimated item difficulties. Our results show that many benchmarks implicitly overweight certain difficulty regions, so their leaderboard conclusions should be interpreted with caution; others leave substantial portions of the difficulty axis uncovered, limiting their ability to monitor the model improvement process. Finally, we turn diagnosis into action by introducing BenchFill, a simple difficulty-targeted item rewriting algorithm that effectively reduces local coverage gaps.

Large Language Models (LLMs) have demonstrated remarkable capabilities across various domains. Math Word Problems (MWPs) serve as a crucial benchmark for evaluating LLMs' reasoning abilities. While most research primarily focuses on improving accuracy, it often neglects understanding and addressing the underlying patterns of errors. Current error classification methods rely on static and predefined categories, which limit their ability to capture the full spectrum of error patterns in mathematical reasoning. To enable systematic error analysis, we collect error samples from 15 different LLMs of varying sizes across four distinct MWP datasets using multiple sampling strategies. Based on this extensive collection, we introduce MWPES-300K, a comprehensive dataset containing 304,865 error samples that cover diverse error patterns and reasoning paths. To reduce human bias and enable fine-grained analysis of error patterns, we propose a novel framework for automated dynamic error classification in mathematical reasoning. Experimental results demonstrate that dataset characteristics significantly shape error patterns, which evolve from basic to complex manifestations as model capabilities increase. With deeper insights into error patterns, we propose error-aware prompting that incorporates common error patterns as explicit guidance, leading to significant improvements in mathematical reasoning performance.

Grammatical Error Detection (GED) methods rely heavily on human annotated error corpora. However, these annotations are unavailable in many low-resource languages. In this paper, we investigate GED in this context. Leveraging the zero-shot cross-lingual transfer capabilities of multilingual pre-trained language models, we train a model using data from a diverse set of languages to generate synthetic errors in other languages. These synthetic error corpora are then used to train a GED model. Specifically we propose a two-stage fine-tuning pipeline where the GED model is first fine-tuned on multilingual synthetic data from target languages followed by fine-tuning on human-annotated GED corpora from source languages. This approach outperforms current state-of-the-art annotation-free GED methods. We also analyse the errors produced by our method and other strong baselines, finding that our approach produces errors that are more diverse and more similar to human errors.

Training dialog policies for speech-based virtual assistants requires a plethora of conversational data. The data collection phase is often expensive and time consuming due to human involvement. To address this issue, a common solution is to build user simulators for data generation. For the successful deployment of the trained policies into real world domains, it is vital that the user simulator mimics realistic conditions. In particular, speech-based assistants are heavily affected by automatic speech recognition and language understanding errors, hence the user simulator should be able to simulate similar errors. In this paper, we review the existing error simulation methods that induce errors at audio, phoneme, text, or semantic level; and conduct detailed comparisons between the audio-level and text-level methods. In the process, we improve the existing text-level method by introducing confidence score prediction and out-of-vocabulary word mapping. We also explore the impact of audio-level and text-level methods on learning a simple clarification dialog policy to recover from errors to provide insight on future improvement for both approaches.

We present a method for classifying syntactic errors in learner language, namely errors whose correction alters the morphosyntactic structure of a sentence. The methodology builds on the established Universal Dependencies syntactic representation scheme, and provides complementary information to other error-classification systems. Unlike existing error classification methods, our method is applicable across languages, which we showcase by producing a detailed picture of syntactic errors in learner English and learner Russian. We further demonstrate the utility of the methodology for analyzing the outputs of leading Grammatical Error Correction (GEC) systems.

Recent advancements in language modeling have enabled the translation of natural language into code, and the use of execution feedback to improve code generation. However, these methods often rely heavily on pre-existing test cases, which may not always be available or comprehensive. In this work, we propose a novel approach that concurrently trains a code generation model and a test generation model, utilizing execution feedback to refine and enhance the performance of both. We introduce two strategies for test and code data augmentation and a new scoring function for code and test ranking. We experiment on the APPS dataset and demonstrate that our approach can effectively generate and augment test cases, filter and synthesize correct code solutions, and rank the quality of generated code and tests. The results demonstrate that our models, when iteratively trained with an increasing number of test cases and code solutions, outperform those trained on the original dataset.

The emergence of generative pre-trained models has facilitated the synthesis of high-quality text, but it has also posed challenges in identifying factual errors in the generated text. In particular: (1) A wider range of tasks now face an increasing risk of containing factual errors when handled by generative models. (2) Generated texts tend to be lengthy and lack a clearly defined granularity for individual facts. (3) There is a scarcity of explicit evidence available during the process of fact checking. With the above challenges in mind, in this paper, we propose FacTool, a task and domain agnostic framework for detecting factual errors of texts generated by large language models (e.g., ChatGPT). Experiments on four different tasks (knowledge-based QA, code generation, mathematical reasoning, and scientific literature review) show the efficacy of the proposed method. We release the code of FacTool associated with ChatGPT plugin interface at https://github.com/GAIR-NLP/factool .

Grammatical error correction (GEC) is the task of correcting typos, spelling, punctuation and grammatical issues in text. Approaching the problem as a sequence-to-sequence task, we compare the use of a common subword unit vocabulary and byte-level encoding. Initial synthetic training data is created using an error-generating pipeline, and used for finetuning two subword-level models and one byte-level model. Models are then finetuned further on hand-corrected error corpora, including texts written by children, university students, dyslexic and second-language writers, and evaluated over different error types and origins. We show that a byte-level model enables higher correction quality than a subword approach, not only for simple spelling errors, but also for more complex semantic, stylistic and grammatical issues. In particular, initial training on synthetic corpora followed by finetuning on a relatively small parallel corpus of real-world errors helps the byte-level model correct a wide range of commonly occurring errors. Our experiments are run for the Icelandic language but should hold for other similar languages, particularly morphologically rich ones.

Ambient AI scribes draft clinical notes, and published audits find their dominant error is omission: information the encounter established that the note fails to record. The standard check is an LLM judge: a second model reads the note against the transcript and flags problems. We ask whether judges detect omissions. Public corpora cannot supply the answer key: their clinician reference notes and transcripts are materially discrepant. Our benchmark has 500 single-error note pairs from audited fact sheets, 298 with a named fact certainly absent and 202 added-or-altered controls. Across eight judge designs, paired discrimination (the flawed note below its clean twin, 0.5 a coin flip) reads 0.79-0.94 on added or altered content and 0.50-0.63 on omissions. On single notes, no design flags omissions reliably more often than perfect notes. Wording changes, voting and GEPA prompt optimisation move the operating point without creating usable detection. Restructuring the task recovers it: list the facts the transcript establishes, then check the note for each. Two methods reach it independently and trade off: a per-fact pipeline, and a GEPA-evolved prompt doing the same in one call. The pipeline's flags name the missing fact and its severity at 2.7% false alarms. The single call detects more (36.9% against 24.6%, p=0.002) at 6.2% false alarms and a tenth of the cost per note. A physician author validated 70 items and, where the two routes disagree, sided with the pipeline on 10 of 10 (p=0.002). A second clinician, not an author, graded the severity rubric blind and agrees to within a grade. On real vendor notes from a companion census no benchmark threshold transfers, but the re-calibrated single call detects more than the best of the eight at half its false-alarm rate. Omissions whose fact is restated elsewhere defeat both routes. We release the benchmark, prompts and judgements.

In this paper, we propose Test-Time Training, a general approach for improving the performance of predictive models when training and test data come from different distributions. We turn a single unlabeled test sample into a self-supervised learning problem, on which we update the model parameters before making a prediction. This also extends naturally to data in an online stream. Our simple approach leads to improvements on diverse image classification benchmarks aimed at evaluating robustness to distribution shifts.

Ambient AI scribes draft clinical notes under the reassurance that a clinician signs every note. We audited three commercial AI scribes on the same 142 consultations: 565 notes from recorded UK primary-care and US ambulatory encounters plus authored scenarios. Twelve discovery passes proposed 13,678 candidate errors; the 5,898 clearing an importance filter went to an adversarial panel of two models from different families, each told to refute what it could, and 618 survived. One note in three (31.3% [27.0, 35.6]) carries a verified failure, concentrated in allergy and medication information, invented patient identity, and history written up as examination on telephone consultations that can contain none. No product was given a patient record; setting aside the two classes a record would have prefilled, invented identity and dates, the rate is 24.8% [20.8, 29.0]. One failure mode did not fit our scheme, drawn from published scribe-error taxonomies: a treatment the clinician retracts, recorded as delivered care. Two clinicians adjudicated blind, disjoint samples: a physician author upheld 20 of 21 findings (95.2% [77.3, 99.2]) and an independent clinician, not an author, 12 of 12 ([75.8, 100]); both judged every sampled refusal genuine. A failure rate depends on the instrument as much as the scribes. With model, evidence and settings fixed, the review instruction alone moves the share of candidates verified from 9.3% to 79.0%, and the reviewing family moves it too: alone at that instruction the gentler flags 54.8% of notes against 27.8%. Between 28% and 97% of sampled notes carry a failure depending on the standard. Published audits disagree among themselves by a margin instrument differences alone can produce: omission is 54-86% of their errors against our 23.1%. We release all 618 findings with transcript-side evidence, every prompt and model version, and the re-runnable pipeline.