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

Tool-augmented general task solving on AgentDojo

94.8Success Rate

Pass@N (oracle)

Updated 1mo ago

Evaluation Results

Method	Links
Pass@N (oracle) 2026.06		94.8
Greedy Decoding 2026.06		94.8
Progress Advantage 2026.06		91.8
Mean-of-N 2026.06		89
ThinkPRM-14B 2026.06		88.7
WildReward-8B 2026.06		86.6
DeepConf Tail 2026.06		86.6
DeepConf B10 2026.06		86.6
ThinkPRM-7B 2026.06		85.6
Self-Certainty 2026.06		85.6
Pass@N (oracle) 2026.06		48.5
Greedy Decoding 2026.06		48.5
ThinkPRM-14B 2026.06		44.3
WildReward-8B 2026.06		43.3
Progress Advantage 2026.06		43.3
Mean-of-N 2026.06		38.8
ThinkPRM-7B 2026.06		37.1
DeepConf Tail 2026.06		35.1
Self-Certainty 2026.06		33
DeepConf B10 2026.06		30.9