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

Multi-step Mathematical Reasoning on MMLU-Pro (held-out test)

83.67Accuracy

LLM-Debate

53.187661.101369.01576.9287May 27, 2026
Updated 6d ago

Evaluation Results

MethodLinks
2026.05
83.67553,410,000
2026.05
82.6697,280,000
2026.05
82.59331,480,000
2026.05
79.7411,040,000
2026.05
79.3232,870,000
2026.05
76.1245,690,000
2026.05
74.7114,460,000
2026.05
74.2320,620,000
2026.05
73.95132,450,000
2026.05
63.556,430,000
2026.05
62.195,470,000
2026.05
59.032,530,000
2026.05
55.723,180,000
2026.05
54.362,790,000