Competition-level Mathematics and Science Reasoning on OlympiadBench (Accuracy)
22.53AccuracyRepeated Sampling
Evaluation Results
| Method | Links | |
|---|---|---|
| Repeated SamplingModel=Qwen2.5-3B-Instruct, Decoding=Majority Voting2025.10 | 22.53 | |
| GUIDEDSAMPLINGModel=Qwen2.5-3B-Instruct, Decoding=Majority Voting2025.10 | 21.87 | |
| GUIDEDSAMPLINGModel=Llama-3.2-3B-Instruct, Decoding=Majority Voting2025.10 | 18.35 | |
| Repeated SamplingModel=Llama-3.2-3B-Instruct, Decoding=Majority Voting2025.10 | 17.47 | |
| Tree-of-thoughtModel=Qwen2.5-3B-Instruct, Decoding=Majority Voting2025.10 | 15.27 | |
| Tree-of-thoughtModel=Llama-3.2-3B-Instruct, Decoding=Majority Voting2025.10 | 12.75 |