Approximation of Log-Partition Function in Policy Mirror Descent Induces Implicit Regularization for LLM Post-Training
Open paper
Benchmark
Model
Rank
Results
mathematical-reasoning-on-aime24
PMD-mean (Qwen3-30B-A3B-Base, τ=0.1)
#80
Acc: 50.83
mathematical-reasoning-on-aime-2025
PMD-mean (Qwen3-30B-A3B-Base, τ=0.1)
#103
Accuracy: 37.19
Rank counts only results with a code link.