Mathematical Reasoning on MATH500 Llama3-8B (first 100 examples)
39AccuracyToT (BFS)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ToT (BFS)Backbone=Llama3-8B-Instruct, Search Algorithm=BFS, Components=User-registered search algorithm, Iterations=10, Branching Factor=3, Temperature=0.7–0.82026.02 | 39 | 1.53 | 2.8 | 14.7 | |
| ReST (MCTS)Backbone=Llama3-8B-Instruct, Search Algorithm=MCTS, Components=built-in, Iterations=10, Branching Factor=3, Temperature=0.7–0.82026.02 | 37 | 2.24 | 4 | 26 | |
| RAP (MCTS)Backbone=Llama3-8B base, Search Algorithm=MCTS, Components=User-registered, Iterations=10, Branching Factor=3, Temperature=0.7–0.82026.02 | 18 | 4.47 | 3.6 | 8 | |
| CoTBackbone=Llama3-8B, Components=built-in2026.02 | 17 | 12.9 | 100 | 0.6 |