What AUROC will the best model for Trojan Detection reach on the Final Round dataset for the NeurIPS Trojan Challenge?
2
resolved Nov 13
<=99% AUROC
Chosen
1.5%
apartresearch avatar<=92.5% AUROC
18%
apartresearch avatar<=80% AUROC
1.5%
apartresearch avatar<=98% AUROC
1.5%
apartresearch avatar<=97% AUROC
1.5%
apartresearch avatar<=96% AUROC
1.5%
apartresearch avatar<=95% AUROC
1.5%
apartresearch avatar<=90% AUROC
1.4%

The current leaderboard shows performance on the validation set. When the Final Round phase begins, we will see the results on the test set. The test set is a held-out dataset available from Oct. 16 2022. The Final Round dataset is created by the parallel track at NeurIPS for creating evasive Trojans. Be aware: This market depends on the evasive Trojans' attack performance.

This question will resolve to the value closest to the highest score on the Final Round dataset.

The current high score on the validation set consisting of infected neural networks is 98.2% AUROC.

🏅 Top traders

#NameTotal profit
1Ṁ100
Sort by:
apartresearch avatar
Apart Research

The market resolves when they release the "Final Round" results on this page: https://codalab.lisn.upsaclay.fr/competitions/5951#results

Related markets

How good will the AUROC be at the end of the Moral Uncertainty Research Competition?
Which Proofnik will get the "Trustworthy. ish." badge next?
What will be the most common word we use for processing text with large language models?
If Redwood Research releases an ELK benchmark paper, will I think it's great backchained empirical alignment research?74%
What will the charity with the most cost-effective intervention+region on Givewell's spreadsheet at the end of 2023 do?
Which photo from this photoshoot will score the highest on Photofeeler?
Will anyone post an interesting math/algorithms koan/problem/exercise in the comments of this that I'll spend 8h+ on?31%
What will be the first letter of the Book Review to win the ACX Book Review Contest?
Will anyone post an interesting math/algorithms koan/problem/exercise in the comments of this that I'll spend 1h+on?85%
What will be the most common name at Proof School among the student and faculty body?
What will I put in first place for the Lodestar Award for Best YA this year?
Conditional on Tower producing a qualifying magazine, will a poll of ACX readers show that most of them find it to be of equal or greater quality to Asterisk Magazine on intellectual rigor?11%
Will anyone post an interesting math/algorithms koan/problem/exercise in the comments of this that I'll spend 30+min on?98%
What Papers/Posts/Books on ML-Theory will I invest more than 5 hours in this year?
Which paper will be retract next from major journals?
What questions will I find most challenging to answer?
If a solid neurological study of trans women gets performed, what will the results most resemble?
What percent of those taking rapamycin to slow aging will comment a positive or neutral review?54%
Will I think it would have been ex-post better to try to get a job at Anthropic instead of working at Redwood Research?34%
What is the best opening move in expert Minesweeper?