Rectifying LLM Thought from Lens of Optimization
Open paper
Benchmark
Model
Rank
Results
mathematical-reasoning-on-math-500
DeepSeek-R1-Distill-Qwen-1.5B with PPO+RePro
#63
Accuracy: 87.7
mathematical-reasoning-on-aime24
DeepSeek-R1-Distill-Qwen-1.5B with PPO+RePro
#98
Acc: 36.3
Rank counts only results with a code link.