Toward Self-Improvement of LLMs via Imagination, Searching, and Criticizing

Open paper
Benchmark Model Rank Results
gsm8k-on-gsm8kAlphaLLM (with MCTS)#1Accuracy: 92
arithmetic-reasoning-on-gsm8kAlphaLLM (with MCTS)#56Accuracy: 92Parameters (Billion): 70
math-word-problem-solving-on-mathAlphaLLM (with MCTS)#81Accuracy: 51

Rank counts only results with a code link.