List your company

Mathematical Reasoning on MATH (test) (Accuracy + Token Count)

94.16AccuracyInjectRLOpt
27.641644.910862.1879.4492Feb 16, 2025Apr 17, 2025Jun 16, 2025Aug 15, 2025Oct 14, 2025Dec 13, 2025Feb 12, 2026

Evaluation results

MethodLinks
InjectRLOptbackbone=Qwen3-32B2026.0294.16—
Baselinebackbone=Qwen3-32B2026.0293.84—
InjectCorrectbackbone=Qwen3-32B2026.0293.8—
Baselinebackbone=Qwen3-14B2026.0293.64—
InjectRLOptbackbone=Qwen3-14B2026.0293.32—
InjectRLOptbackbone=Qwen3-8B2026.0293.28—
InjectCorrectbackbone=Qwen3-14B2026.0292.96—
Baselinebackbone=Qwen3-4B2026.0292.88—
InjectRLOptbackbone=Qwen3-4B2026.0292.88—
Baselinebackbone=Qwen3-8B2026.0292.52—
InjectCorrectbackbone=Qwen3-4B2026.0292.4—
InjectCorrectbackbone=Qwen3-8B2026.0292.33—
MCTS w/ FETCHsearch_strategy=MCTS, variant=FETCH2025.024022.9
Beam Search w/ FETCHsearch_strategy=Beam Search, variant=FETCH2025.0239.67.04
MCTS w/ Var Reducesearch_strategy=MCTS, variant=Var Reduce2025.0239.466.6
BFS w/ Var Reducesearch_strategy=BFS, variant=Var Reduce2025.0238.87.63
Beam Search w/ Var Reducesearch_strategy=Beam Search, variant=Var Reduce2025.0238.67.45
BFS w/ FETCHsearch_strategy=BFS, variant=FETCH2025.0238.43.04
Beam Search w/ State Mergesearch_strategy=Beam Search, variant=State Merge2025.0238.46.36
Tree Search w/ FETCHsearch_strategy=Tree Search, variant=FETCH2025.0238.42
BFSsearch_strategy=BFS2025.0238.26.51
BFS w/ State Mergesearch_strategy=BFS, variant=State Merge2025.02382.89
Tree Search w/ Var Reducesearch_strategy=Tree Search, variant=Var Reduce2025.02383.53
MCTS w/ State Mergesearch_strategy=MCTS, variant=State Merge2025.023820.3
Beam Searchsearch_strategy=Beam Search2025.0237.86.5
Tree Search w/ State Mergesearch_strategy=Tree Search, variant=State Merge2025.0237.81.58
MCTSsearch_strategy=MCTS2025.0237.459
Weighted Votingsearch_strategy=Weighted Voting2025.0237.22.44
Tree Searchsearch_strategy=Tree Search2025.0237.22.68
Best-of-Nsearch_strategy=Best-of-N2025.0235.22.44
Self-Consistencysearch_strategy=Self-Consistency2025.02342.44
Greedy Decodingsearch_strategy=Greedy Decoding2025.0230.20.25