Mathematical Reasoning on MATH (test) (Accuracy + Token Count)
94.16AccuracyInjectRLOpt
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| InjectRLOptbackbone=Qwen3-32B2026.02 | 94.16 | — | |
| Baselinebackbone=Qwen3-32B2026.02 | 93.84 | — | |
| InjectCorrectbackbone=Qwen3-32B2026.02 | 93.8 | — | |
| Baselinebackbone=Qwen3-14B2026.02 | 93.64 | — | |
| InjectRLOptbackbone=Qwen3-14B2026.02 | 93.32 | — | |
| InjectRLOptbackbone=Qwen3-8B2026.02 | 93.28 | — | |
| InjectCorrectbackbone=Qwen3-14B2026.02 | 92.96 | — | |
| Baselinebackbone=Qwen3-4B2026.02 | 92.88 | — | |
| InjectRLOptbackbone=Qwen3-4B2026.02 | 92.88 | — | |
| Baselinebackbone=Qwen3-8B2026.02 | 92.52 | — | |
| InjectCorrectbackbone=Qwen3-4B2026.02 | 92.4 | — | |
| InjectCorrectbackbone=Qwen3-8B2026.02 | 92.33 | — | |
| MCTS w/ FETCHsearch_strategy=MCTS, variant=FETCH2025.02 | 40 | 22.9 | |
| Beam Search w/ FETCHsearch_strategy=Beam Search, variant=FETCH2025.02 | 39.6 | 7.04 | |
| MCTS w/ Var Reducesearch_strategy=MCTS, variant=Var Reduce2025.02 | 39.4 | 66.6 | |
| BFS w/ Var Reducesearch_strategy=BFS, variant=Var Reduce2025.02 | 38.8 | 7.63 | |
| Beam Search w/ Var Reducesearch_strategy=Beam Search, variant=Var Reduce2025.02 | 38.6 | 7.45 | |
| BFS w/ FETCHsearch_strategy=BFS, variant=FETCH2025.02 | 38.4 | 3.04 | |
| Beam Search w/ State Mergesearch_strategy=Beam Search, variant=State Merge2025.02 | 38.4 | 6.36 | |
| Tree Search w/ FETCHsearch_strategy=Tree Search, variant=FETCH2025.02 | 38.4 | 2 | |
| BFSsearch_strategy=BFS2025.02 | 38.2 | 6.51 | |
| BFS w/ State Mergesearch_strategy=BFS, variant=State Merge2025.02 | 38 | 2.89 | |
| Tree Search w/ Var Reducesearch_strategy=Tree Search, variant=Var Reduce2025.02 | 38 | 3.53 | |
| MCTS w/ State Mergesearch_strategy=MCTS, variant=State Merge2025.02 | 38 | 20.3 | |
| Beam Searchsearch_strategy=Beam Search2025.02 | 37.8 | 6.5 | |
| Tree Search w/ State Mergesearch_strategy=Tree Search, variant=State Merge2025.02 | 37.8 | 1.58 | |
| MCTSsearch_strategy=MCTS2025.02 | 37.4 | 59 | |
| Weighted Votingsearch_strategy=Weighted Voting2025.02 | 37.2 | 2.44 | |
| Tree Searchsearch_strategy=Tree Search2025.02 | 37.2 | 2.68 | |
| Best-of-Nsearch_strategy=Best-of-N2025.02 | 35.2 | 2.44 | |
| Self-Consistencysearch_strategy=Self-Consistency2025.02 | 34 | 2.44 | |
| Greedy Decodingsearch_strategy=Greedy Decoding2025.02 | 30.2 | 0.25 |