List your company
ResearchFormerly WizWand

Math Reasoning on AMC (avg@8 accuracy)

47.29Avg@8 AccuracyJURY-RL
15.965224.097632.2340.3624Apr 28, 2026
Updated 6mo ago

Evaluation results

MethodLinks
JURY-RLBackbone=Qwen2.5-7B2026.0447.29
GT-RewardBackbone=Qwen2.5-7B2026.0446.54
LLM-KDBackbone=Qwen2.5-7B2026.0444.28
LLM-as-a-JudgeBackbone=Qwen2.5-7B2026.0441.11
CoRewardBackbone=Qwen2.5-7B2026.0440.81
Self-CertaintyBackbone=Qwen2.5-7B2026.0440.51
EntropyBackbone=Qwen2.5-7B2026.0440.51
Majority-VotingBackbone=Qwen2.5-7B2026.0438.7
LLM-KDBackbone=Qwen3-1.7B-Base2026.0435.69
JURY-RLBackbone=Qwen3-1.7B-Base2026.0434.19
LLM-as-a-JudgeBackbone=Qwen3-1.7B-Base2026.0433.89
CoRewardBackbone=Qwen3-1.7B-Base2026.0433.13
EntropyBackbone=Qwen3-1.7B-Base2026.0431.17
GT-RewardBackbone=Qwen3-1.7B-Base2026.0430.87
Majority-VotingBackbone=Qwen3-1.7B-Base2026.0429.82
Self-CertaintyBackbone=Qwen3-1.7B-Base2026.0427.41
JURY-RLBackbone=Llama-3.2-3B-Instruct2026.0426.05
LLM-KDBackbone=Llama-3.2-3B-Instruct2026.0424.55
GT-RewardBackbone=Llama-3.2-3B-Instruct2026.0422.89
Before RLBackbone=Qwen2.5-7B2026.0422.44
Majority-VotingBackbone=Llama-3.2-3B-Instruct2026.0421.69
LLM-as-a-JudgeBackbone=Llama-3.2-3B-Instruct2026.0420.93
Before RLBackbone=Qwen3-1.7B-Base2026.0420.78
CoRewardBackbone=Llama-3.2-3B-Instruct2026.0419.58
Self-CertaintyBackbone=Llama-3.2-3B-Instruct2026.0417.62
Before RLBackbone=Llama-3.2-3B-Instruct2026.0417.47
EntropyBackbone=Llama-3.2-3B-Instruct2026.0417.17