DeepFunding/originality-r3-data

Deep Funding Originality: Round 3: data

All files are in the dataset DeepFunding/originality-r3-data, which is linked from the competition Deep Funding Originality: Round 3.

File	Rows	What it holds
repos.csv	98	The competition repositories, spelled as the markets spell them, with GitHub URL, market bundle and Seer market.
sample_submission.csv	98	A correctly formatted submission. Every value is 0.5: a format example, not a forecast.
r1_juror_ratings.csv	51	Every juror rating from Deep Funding's first originality round (Originality R1): repository, the name the juror used, the 0–100 rating and the juror's note, if any.
public_leaderboard_targets.csv	33	The public leaderboard's answers: each R1-rated repository's average rating divided by 100, with the number of ratings.
market_prices.csv	98	UP and DOWN prices of each repository's Round 3 originality market on deep.seer.pm at 2026-10-16 10:55 UTC.
metric.py		The leaderboard's scorer. Run it to check a submission and score it against the public targets.

Notes

• Originality R1 used two older names, mapped here onto the market names: ethereum/solidity → argotorg/solidity (the repository moved) and account-abstraction → eth-infinitism/account-abstraction.
• Each Round 3 market pays UP in proportion to the repository's final average originality, so the UP price is the market's estimate of it. Prices are in units of the market's repo token. UP and DOWN trade in separate pools, and part of the value sits in the market's Invalid outcome, so UP + DOWN is not exactly 1 (0.85 to 1.05 in this snapshot); up_share = UP / (UP + DOWN) rescales the pair to sum to 1. Live prices are on deep.seer.pm.
• The private leaderboard's answers are the Round 3 jurors' averages. They are not here because they do not exist yet.

License

r1_juror_ratings.csv and public_leaderboard_targets.csv come from deepfunding/oss-evals (MIT License, © 2026 Deep Funding). The full license text is in LICENSE.