List your company
ResearchFormerly WizWand

Math Reasoning on AIME 2024 (avg@16 accuracy)

18.33Accuracy (avg@16)GT-Reward
1.435.817510.20514.5925Apr 28, 2026
Updated 6mo ago

Evaluation results

MethodLinks
GT-RewardBackbone=Qwen2.5-7B2026.0418.33
JURY-RLBackbone=Qwen2.5-7B2026.0413.75
LLM-KDBackbone=Qwen2.5-7B2026.0412.08
Majority-VotingBackbone=Qwen2.5-7B2026.0411.25
CoRewardBackbone=Qwen2.5-7B2026.0411.25
EntropyBackbone=Qwen2.5-7B2026.0411.04
GT-RewardBackbone=Llama-3.2-3B-Instruct2026.0410.62
LLM-as-a-JudgeBackbone=Qwen2.5-7B2026.0410.42
LLM-KDBackbone=Llama-3.2-3B-Instruct2026.0410.21
Self-CertaintyBackbone=Qwen2.5-7B2026.049.79
LLM-as-a-JudgeBackbone=Llama-3.2-3B-Instruct2026.049.38
JURY-RLBackbone=Llama-3.2-3B-Instruct2026.049.17
Majority-VotingBackbone=Llama-3.2-3B-Instruct2026.048.33
JURY-RLBackbone=Qwen3-1.7B-Base2026.047.71
CoRewardBackbone=Llama-3.2-3B-Instruct2026.047.71
GT-RewardBackbone=Qwen3-1.7B-Base2026.046.88
CoRewardBackbone=Qwen3-1.7B-Base2026.046.88
Before RLBackbone=Llama-3.2-3B-Instruct2026.046.88
LLM-KDBackbone=Qwen3-1.7B-Base2026.046.67
EntropyBackbone=Qwen3-1.7B-Base2026.046.67
LLM-as-a-JudgeBackbone=Qwen3-1.7B-Base2026.045.83
Before RLBackbone=Qwen2.5-7B2026.044.38
Self-CertaintyBackbone=Qwen3-1.7B-Base2026.043.54
EntropyBackbone=Llama-3.2-3B-Instruct2026.043.54
Before RLBackbone=Qwen3-1.7B-Base2026.043.12
Self-CertaintyBackbone=Llama-3.2-3B-Instruct2026.042.71
Majority-VotingBackbone=Qwen3-1.7B-Base2026.042.08