DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Search
Open paper
Benchmark
Model
Rank
Results
automated-theorem-proving-on-minif2f-test
DeepSeek-Prover-V1.5
#3
cumulative: 63.5
Pass@32: 50.0
Pass@64: 50.7
ITP: Lean
Rank counts only results with a code link.