Share your thoughts, 1 month free Claude Pro on usSee more

Code on CRUX

55.08Accuracy @5

CoReward

Updated 1mo ago

Evaluation Results

Method	Links
CoReward 2026.04		55.08	-
JURY-RL 2026.04		54.37	-
GT-Reward 2026.04		53.67	-
LLM-KD 2026.04		53.57	-
Self-Certainty 2026.04		53.18	-
Majority-Voting 2026.04		52.2	-
LLM-as-a-Judge 2026.04		51.9	-
Entropy 2026.04		51.08	-
LLM-as-a-Judge 2026.04		34.98	-
CoReward 2026.04		34.1	-
JURY-RL 2026.04		33.98	-
LLM-KD 2026.04		33.8	-
GT-Reward 2026.04		33.73	-
GT-Reward 2026.04		32.48	-
Majority-Voting 2026.04		32	-
JURY-RL 2026.04		31.6	-
LLM-as-a-Judge 2026.04		31.45	-
LLM-KD 2026.04		31.2	-
CoReward 2026.04		30.87	-
Majority-Voting 2026.04		30.83	-
Entropy 2026.04		30.38	-
Before RL 2026.04		27.38	-
Entropy 2026.04		26.3	-
Before RL 2026.04		24.7	-
Self-Certainty 2026.04		24.55	-
Self-Certainty 2026.04		18.73	-
Before RL 2026.04		7.52	-
Qwen3-14B-Base 2025.08		-	63.9
Qwen2.5-14B-Instruct-1M 2025.08		-	66.4
Qwen2.5-7B-Instruct-1M 2025.08		-	55.6
NPG-Muse-7B 2025.08		-	59
Qwen3-8B-Base 2025.08		-	54.1
NPG-Muse-8B 2025.08		-	65.3