Math Reasoning on AIME, MATH500, GSM8K, AMC (test)
64.04Pass@k ScoreJURY-RL
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| JURY-RLBackbone=Qwen2.5-7B2026.04 | 64.04 | 48.13 | 1.56 | 10.05 | |
| GT-RewardBackbone=Qwen2.5-7B2026.04 | 62.48 | 46.6 | — | — | |
| JURY-RLBackbone=Qwen3-1.7B-Base2026.04 | 59.41 | 41.57 | 4.05 | 9.06 | |
| GT-RewardBackbone=Qwen3-1.7B-Base2026.04 | 55.36 | 39.25 | — | — | |
| LLM-as-a-judgeBackbone=Qwen2.5-7B2026.04 | 53.99 | 44.13 | — | — | |
| LLM-as-a-judgeBackbone=Qwen3-1.7B-Base2026.04 | 50.35 | 37.6 | — | — | |
| JURY-RLBackbone=Llama-3.2-3B-Instruct2026.04 | 48.48 | 34.1 | 3.02 | 2.55 | |
| LLM-as-a-judgeBackbone=Llama-3.2-3B-Instruct2026.04 | 45.93 | 30.65 | — | — | |
| GT-RewardBackbone=Llama-3.2-3B-Instruct2026.04 | 45.46 | 32.19 | — | — |