Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Formal Mathematics on Lean-Workbook 2,500-problem (test)
Loading...
57.1
PCR (%)
TRI (Full)
32.66
39.005
45.35
51.695
Jun 3, 2026
PCR (%)
Token/Probability Metric
Updated 1mo ago
Evaluation Results
Method
Method
Links
PCR (%)
Token/Probability Metric
TRI (Full)
Fine-tuning Protocol=S...
2026.06
57.1
1,268
TRI (SFT + DPO)
Fine-tuning Protocol=S...
2026.06
55.4
1,267
InternLM-StepProver + CoT-SC
Backbone=InternLM-Step...
2026.06
51.2
15,896
TRI (SFT only)
Fine-tuning Protocol=S...
2026.06
49.7
1,534
InternLM-StepProver + CoT
Backbone=InternLM-Step...
2026.06
44.9
1,987
Qwen2.5-72B + CoT-SC
Backbone=Qwen2.5-72B,...
2026.06
43.1
29,472
Qwen2.5-72B + CoT
Backbone=Qwen2.5-72B,...
2026.06
38.2
1,842
Llama-3.1-70B + ToT
Backbone=Llama-3.1-70B...
2026.06
37.4
9,561
Llama-3.1-70B + CoT
Backbone=Llama-3.1-70B...
2026.06
33.6
1,913
Feedback
Search any
task
Search any
task