Math Reasoning on AIME 24 (pass@16, Average)
36.67Pass@16JURY-RL
Evaluation results
| Method | Links | ||
|---|---|---|---|
| JURY-RLBackbone=Qwen2.5-7B2026.04 | 36.67 | 64.04 | |
| GT-RewardBackbone=Qwen2.5-7B2026.04 | 33.33 | 62.48 | |
| LLM-KDBackbone=Qwen2.5-7B2026.04 | 33.33 | 63.54 | |
| Majority-VotingBackbone=Qwen2.5-7B2026.04 | 33.33 | 61.52 | |
| JURY-RLBackbone=Qwen3-1.7B-Base2026.04 | 30 | 59.41 | |
| JURY-RLBackbone=Llama-3.2-3B-Instruct2026.04 | 30 | 48.48 | |
| Before RLBackbone=Qwen2.5-7B2026.04 | 30 | 56.77 | |
| EntropyBackbone=Qwen2.5-7B2026.04 | 30 | 62.62 | |
| CoRewardBackbone=Qwen2.5-7B2026.04 | 30 | 59.99 | |
| GT-RewardBackbone=Qwen3-1.7B-Base2026.04 | 26.67 | 55.36 | |
| CoRewardBackbone=Qwen3-1.7B-Base2026.04 | 26.67 | 53.4 | |
| LLM-as-a-JudgeBackbone=Llama-3.2-3B-Instruct2026.04 | 26.67 | 45.93 | |
| Self-CertaintyBackbone=Qwen3-1.7B-Base2026.04 | 23.33 | 55.17 | |
| Majority-VotingBackbone=Qwen3-1.7B-Base2026.04 | 23.33 | 52.51 | |
| Before RLBackbone=Llama-3.2-3B-Instruct2026.04 | 23.33 | 46.26 | |
| GT-RewardBackbone=Llama-3.2-3B-Instruct2026.04 | 23.33 | 45.46 | |
| LLM-KDBackbone=Llama-3.2-3B-Instruct2026.04 | 23.33 | 46.21 | |
| CoRewardBackbone=Llama-3.2-3B-Instruct2026.04 | 23.33 | 43.51 | |
| Self-CertaintyBackbone=Qwen2.5-7B2026.04 | 23.33 | 60.33 | |
| LLM-KDBackbone=Qwen3-1.7B-Base2026.04 | 20 | 53.51 | |
| EntropyBackbone=Qwen3-1.7B-Base2026.04 | 20 | 54.86 | |
| LLM-as-a-JudgeBackbone=Qwen2.5-7B2026.04 | 20 | 53.99 | |
| Before RLBackbone=Qwen3-1.7B-Base2026.04 | 16.67 | 49.59 | |
| LLM-as-a-JudgeBackbone=Qwen3-1.7B-Base2026.04 | 16.67 | 50.35 | |
| EntropyBackbone=Llama-3.2-3B-Instruct2026.04 | 16.67 | 42.77 | |
| Majority-VotingBackbone=Llama-3.2-3B-Instruct2026.04 | 16.67 | 42.2 | |
| Self-CertaintyBackbone=Llama-3.2-3B-Instruct2026.04 | 13.33 | 42.12 |