List your company
ResearchFormerly WizWand

Mathematical Reasoning on MATH 500 (Mean@1)

0.992Mean@1FlashMLA
0.105920.335960.5660.79604Jun 4, 2025Jul 19, 2025Sep 3, 2025Oct 19, 2025Dec 3, 2025Jan 18, 2026Mar 5, 2026
Updated 8mo ago

Evaluation results

MethodLinks
FlashMLABackbone=LongCat-Flash-thinking, Precision=BF162026.020.992
FlashMLABackbone=DeepSeek-V3.1, Precision=BF162026.020.988
SnapMLABackbone=LongCat-Flash-thinking, Precision=FP82026.020.988
SnapMLABackbone=DeepSeek-V3.1, Precision=FP82026.020.982
R1-Distill-Qwen-7B-R-TAP2026.030.927
CyclicReflexBackbone=Qwen3-14B, Decoding Strategy=CyclicReflex2025.060.9
R1-Distill-Qwen-7B @ 8kTraining iterations=8k2026.030.881
OriginalBackbone=Qwen3-14B, Decoding Strategy=Original2025.060.88
TIPBackbone=Qwen3-14B, Decoding Strategy=TIP2025.060.88
Oat-Zero-7B-R-TAP2026.030.872
CyclicReflexBackbone=Qwen3-8B, Decoding Strategy=CyclicReflex2025.060.86
TIPBackbone=Qwen3-8B, Decoding Strategy=TIP2025.060.84
PRIME-Zero-7B2026.030.838
Qwen2.5-Math-7B-Instruct2026.030.836
R1-Distill-Qwen-1.5B-R-TAP @ 8kTraining iterations=8k2026.030.835
OriginalBackbone=Qwen3-8B, Decoding Strategy=Original2025.060.83
OpenReasoner-Zero-7B @ 8kTraining iterations=8k2026.030.824
DiSCTTModel=Qwen-2.5-7B-Instruct2026.030.822
Oat-Zero-1.5B-R-TAP2026.030.8
Oat-Zero-7B2026.030.8
OpenReasoner-Zero-7B @ 3kTraining iterations=3k2026.030.792
SimpleRL-Zero-7B2026.030.782
R1-Distill-Qwen-1.5B @ 8kTraining iterations=8k2026.030.774
DiSCTTModel=Qwen-3-4B-Base2026.030.752
Qwen2.5-Math-1.5B-Instruct2026.030.742
Oat-Zero-1.5B2026.030.742
TTRLModel=Qwen-2.5-7B-Instruct2026.030.742
EVOL-RLModel=Qwen-2.5-7B-Instruct2026.030.734
Qwen2.5-Math-7B2026.030.69
CyclicReflexBackbone=Qwen3-4B, Decoding Strategy=CyclicReflex2025.060.68
DiSCTTModel=Qwen-3-1.7B-Base2026.030.67
EVOL-RLModel=Qwen-3-4B-Base2026.030.646
TIPBackbone=Qwen3-4B, Decoding Strategy=TIP2025.060.63
Qwen2.5-Math-1.5B2026.030.618
TTRLModel=Qwen-3-4B-Base2026.030.606
R1-Distill-Qwen-7B @ 3kTraining iterations=3k2026.030.601
OriginalBackbone=Qwen3-4B, Decoding Strategy=Original2025.060.6
BaseModel=Qwen-2.5-7B-Instruct2026.030.588
DiSCTTModel=LLaMA-3.2-3B-Instruct2026.030.556
TTRLModel=Qwen-3-1.7B-Base2026.030.552
EVOL-RLModel=Qwen-3-1.7B-Base2026.030.538
R1-Distill-Qwen-1.5B @ 3kTraining iterations=3k2026.030.522
BaseModel=Qwen-3-4B-Base2026.030.514
DiSCTTModel=Qwen-2.5-0.5B-Instruct2026.030.498
EVOL-RLModel=LLaMA-3.2-3B-Instruct2026.030.472
BaseModel=Qwen-3-1.7B-Base2026.030.47
TTRLModel=LLaMA-3.2-3B-Instruct2026.030.454
EVOL-RLModel=Qwen-2.5-0.5B-Instruct2026.030.4
DiSCTTModel=LLaMA-3.2-1B-Instruct2026.030.396
TTRLModel=Qwen-2.5-0.5B-Instruct2026.030.372
BaseModel=LLaMA-3.2-3B-Instruct2026.030.357
EVOL-RLModel=LLaMA-3.2-1B-Instruct2026.030.312
TTRLModel=LLaMA-3.2-1B-Instruct2026.030.28
BaseModel=Qwen-2.5-0.5B-Instruct2026.030.276
BaseModel=LLaMA-3.2-1B-Instruct2026.030.14