Math Reasoning on MATH 500 (Acc, ∆Tok)
99.2AccuracyQwen3-Next-80B + THINKBRAKE
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Qwen3-Next-80B + THINKBRAKEBackbone=Qwen3-Next-80B, Decoding Strategy=THINKBRAKE2025.10 | 99.2 | -21.6 | |
| Qwen3-Next-80BBackbone=Qwen3-Next-80B, Decoding Strategy=Base2025.10 | 98.2 | — | |
| Qwen3-32B + THINKBRAKEBackbone=Qwen3-32B, Decoding Strategy=THINKBRAKE2025.10 | 97.2 | -1.4 | |
| Qwen3-32BBackbone=Qwen3-32B, Decoding Strategy=Base2025.10 | 97 | — | |
| Qwen3-14B + THINKBRAKEBackbone=Qwen3-14B, Decoding Strategy=THINKBRAKE2025.10 | 96.9 | -7.7 | |
| Qwen3-14BBackbone=Qwen3-14B, Decoding Strategy=Base2025.10 | 96.8 | — | |
| Qwen3-4B-Thinking + THINKBRAKEBackbone=Qwen3-4B-Thinking, Decoding Strategy=THINKBRAKE2025.10 | 96.6 | -20.4 | |
| Qwen3-4B-Thinking + NoWaitBackbone=Qwen3-4B-Thinking, Decoding Strategy=NoWait2025.10 | 96.4 | -17.9 | |
| Qwen3-4BBackbone=Qwen3-4B, Decoding Strategy=Base2025.10 | 96 | — | |
| Qwen3-32B + NoWaitBackbone=Qwen3-32B, Decoding Strategy=NoWait2025.10 | 95.8 | -20.6 | |
| Qwen3-4B + THINKBRAKEBackbone=Qwen3-4B, Decoding Strategy=THINKBRAKE2025.10 | 95.4 | -14.4 | |
| Qwen3-4B-ThinkingBackbone=Qwen3-4B-Thinking, Decoding Strategy=Base2025.10 | 95.3 | — | |
| Qwen3-14B + NoWaitBackbone=Qwen3-14B, Decoding Strategy=NoWait2025.10 | 95.2 | -25.1 | |
| Qwen3-4B + NoWaitBackbone=Qwen3-4B, Decoding Strategy=NoWait2025.10 | 94.6 | -31.4 | |
| Qwen3-4B-Thinking + DEERBackbone=Qwen3-4B-Thinking, Decoding Strategy=DEER2025.10 | 94.4 | -14.2 | |
| DeepSeek-R1-7BBackbone=DeepSeek-R1-7B, Decoding Strategy=Base2025.10 | 93.8 | — | |
| DeepSeek-R1-7B + THINKBRAKEBackbone=DeepSeek-R1-7B, Decoding Strategy=THINKBRAKE2025.10 | 92 | -23.2 | |
| Qwen3-4B-Thinking + THINKBRAKE-pBackbone=Qwen3-4B-Thinking, Decoding Strategy=THINKBRAKE-p2025.10 | 91.2 | -57.9 | |
| DeepSeek-R1-7B + NoWaitBackbone=DeepSeek-R1-7B, Decoding Strategy=NoWait2025.10 | 91.2 | -31 | |
| Qwen3-32B + THINKBRAKE-pBackbone=Qwen3-32B, Decoding Strategy=THINKBRAKE-p2025.10 | 89.6 | -39.4 | |
| Qwen3-14B + THINKBRAKE-pBackbone=Qwen3-14B, Decoding Strategy=THINKBRAKE-p2025.10 | 89 | -49.8 | |
| Qwen3-4B + THINKBRAKE-pBackbone=Qwen3-4B, Decoding Strategy=THINKBRAKE-p2025.10 | 88.6 | -51.3 | |
| Qwen3-32B + ThinkLessBackbone=Qwen3-32B, Decoding Strategy=ThinkLess2025.10 | 87.6 | -100 | |
| Qwen3-4B-Thinking + Dynasor-CoTBackbone=Qwen3-4B-Thinking, Decoding Strategy=Dynasor-CoT2025.10 | 87.2 | -42.1 | |
| DeepSeek-R1-7B + DEERBackbone=DeepSeek-R1-7B, Decoding Strategy=DEER2025.10 | 87.2 | -44.5 | |
| Qwen3-14B + ThinkLessBackbone=Qwen3-14B, Decoding Strategy=ThinkLess2025.10 | 86.8 | -100 | |
| Qwen3-4B + ThinkLessBackbone=Qwen3-4B, Decoding Strategy=ThinkLess2025.10 | 85.4 | -100 | |
| DeepSeek-R1-7B + THINKBRAKE-pBackbone=DeepSeek-R1-7B, Decoding Strategy=THINKBRAKE-p2025.10 | 81.4 | -60.6 | |
| DeepSeek-R1-7B + ThinkLessBackbone=DeepSeek-R1-7B, Decoding Strategy=ThinkLess2025.10 | 78.4 | -100 | |
| Phi-4-ReasoningBackbone=Phi-4-Reasoning, Decoding Strategy=Base2025.10 | 71.1 | — | |
| Phi-4-Reasoning + NoWaitBackbone=Phi-4-Reasoning, Decoding Strategy=NoWait2025.10 | 70.8 | -17.7 | |
| Phi-4-Reasoning + THINKBRAKEBackbone=Phi-4-Reasoning, Decoding Strategy=THINKBRAKE2025.10 | 70.4 | -17.2 | |
| Phi-4-Reasoning + THINKBRAKE-pBackbone=Phi-4-Reasoning, Decoding Strategy=THINKBRAKE-p2025.10 | 68.8 | -50.4 | |
| Phi-4-Reasoning + ThinkLessBackbone=Phi-4-Reasoning, Decoding Strategy=ThinkLess2025.10 | 66.4 | -100 | |
| DeepSeek-R1-7B + Dynasor-CoTBackbone=DeepSeek-R1-7B, Decoding Strategy=Dynasor-CoT2025.10 | 64.2 | -47.2 | |
| Qwen3-4B-Thinking + ThinkLessBackbone=Qwen3-4B-Thinking, Decoding Strategy=ThinkLess2025.10 | 56.4 | -100 |