List your company
ResearchFormerly WizWand

Code Generation on LiveCodeBench (avg@5 accuracy)

18.37Avg@5 AccuracyMajority-Voting
2.38526.535110.68514.8349Apr 28, 2026
Updated 6mo ago

Evaluation results

MethodLinks
Majority-VotingBackbone=Qwen2.5-7B2026.0418.37
EntropyBackbone=Qwen2.5-7B2026.0415.68
GT-RewardBackbone=Qwen3-1.7B-Base2026.0414.84
JURY-RLBackbone=Qwen2.5-7B2026.0414.69
JURY-RLBackbone=Qwen3-1.7B-Base2026.0414.54
LLM-as-a-JudgeBackbone=Qwen3-1.7B-Base2026.0414.33
Majority-VotingBackbone=Qwen3-1.7B-Base2026.0414.29
CoRewardBackbone=Qwen3-1.7B-Base2026.0414.27
LLM-KDBackbone=Qwen3-1.7B-Base2026.0414.05
GT-RewardBackbone=Qwen2.5-7B2026.0412.78
EntropyBackbone=Qwen3-1.7B-Base2026.0412.64
Self-CertaintyBackbone=Qwen2.5-7B2026.0411.87
LLM-as-a-JudgeBackbone=Qwen2.5-7B2026.0410.58
Self-CertaintyBackbone=Qwen3-1.7B-Base2026.0410.37
CoRewardBackbone=Qwen2.5-7B2026.0410.37
Majority-VotingBackbone=Llama-3.2-3B-Instruct2026.048.11
LLM-KDBackbone=Qwen2.5-7B2026.048.09
LLM-KDBackbone=Llama-3.2-3B-Instruct2026.047.55
GT-RewardBackbone=Llama-3.2-3B-Instruct2026.047.05
JURY-RLBackbone=Llama-3.2-3B-Instruct2026.046.16
Self-CertaintyBackbone=Llama-3.2-3B-Instruct2026.045.57
CoRewardBackbone=Llama-3.2-3B-Instruct2026.045.57
EntropyBackbone=Llama-3.2-3B-Instruct2026.045.38
Before RLBackbone=Qwen3-1.7B-Base2026.044.59
Before RLBackbone=Qwen2.5-7B2026.044.57
LLM-as-a-JudgeBackbone=Llama-3.2-3B-Instruct2026.043.96
Before RLBackbone=Llama-3.2-3B-Instruct2026.043