Degree of Change on PerturbQA K562
88.32AccuracyBoN w/ Full Set
Evaluation Results
| Method | Links | |
|---|---|---|
| BoN w/ Full SetDecoding Strategy=Best-of-N, Reward Model=Full Set2026.03 | 88.32 | |
| Majority VotingDecoding Strategy=Majority Voting2026.03 | 85.1 | |
| BoN w/ Qwen-PRM800KDecoding Strategy=Best-of-N, Reward Model=Qwen-PRM800K2026.03 | 84.87 | |
| Greedy DecodingDecoding Strategy=Greedy2026.03 | 83.47 | |
| BoN w/ Llama-PRM800KDecoding Strategy=Best-of-N, Reward Model=Llama-PRM800K2026.03 | 82.84 | |
| BoN w/ VersaPRMDecoding Strategy=Best-of-N, Reward Model=VersaPRM2026.03 | 80.36 |