Mathematical Reasoning on MATH500 4-shot
42.2AccuracyVanilla
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| VanillaBackbone=Dream-v0-base-7B, Parallel Decode=Vanilla, Block=|B0| = 322026.04 | 42.2 | 9.7 | 256 | |
| DepCapBackbone=LLaDA-1.5, Parallel Decode=CAP-Decoding, Block=DepGA-Block2026.04 | 42.2 | 19.7 | 94.1 | |
| DepGA-BlockBackbone=LLaDA-1.5, Parallel Decode=Confidence, Block=DepGA-Block2026.04 | 41.6 | 18.1 | 100 | |
| ConfidenceBackbone=Dream-v0-base-7B, Parallel Decode=Confidence, Block=|B0| = 642026.04 | 41 | 24.9 | 99.5 | |
| DepGA-BlockBackbone=Dream-v0-base-7B, Parallel Decode=Confidence, Block=DepGA-Block2026.04 | 41 | 24 | 103.2 | |
| ConfidenceBackbone=Dream-v0-base-7B, Parallel Decode=Confidence, Block=|B0| = 162026.04 | 40.8 | 23.7 | 105.5 | |
| DepCapBackbone=Dream-v0-base-7B, Parallel Decode=CAP-Decoding, Block=DepGA-Block2026.04 | 40.8 | 24.4 | 97.2 | |
| DepGA-BlockBackbone=LLaDA-8B-Instruct, Parallel Decode=Confidence, Block=DepGA-Block2026.04 | 40.4 | 19.9 | 101.8 | |
| ConfidenceBackbone=Dream-v0-base-7B, Parallel Decode=Confidence, Block=|B0| = 322026.04 | 40.4 | 24.4 | 102.3 | |
| CAP-DecodingBackbone=Dream-v0-base-7B, Parallel Decode=CAP-Decoding, Block=|B0| = 322026.04 | 40.4 | 24.7 | 95.4 | |
| ConfidenceBackbone=LLaDA-8B-Instruct, Parallel Decode=Confidence, Block=|B0| = 162026.04 | 40.2 | 19.7 | 104.4 | |
| AdaBlockBackbone=LLaDA-8B-Instruct, Parallel Decode=Confidence, Block=AdaBlock2026.04 | 40 | 19.8 | 103.5 | |
| ConfidenceBackbone=LLaDA-1.5, Parallel Decode=Confidence, Block=|B0| = 162026.04 | 40 | 17.8 | 101 | |
| CAP-DecodingBackbone=LLaDA-1.5, Parallel Decode=CAP-Decoding, Block=|B0| = 322026.04 | 40 | 20.2 | 88.8 | |
| DepCapBackbone=LLaDA-8B-Instruct, Parallel Decode=CAP-Decoding, Block=DepGA-Block2026.04 | 39.8 | 22.2 | 94.8 | |
| AdaBlockBackbone=Dream-v0-base-7B, Parallel Decode=Confidence, Block=AdaBlock2026.04 | 39.8 | 24.7 | 101.7 | |
| VanillaBackbone=LLaDA-1.5, Parallel Decode=Vanilla, Block=|B0| = 322026.04 | 39.8 | 7.1 | 256 | |
| ConfidenceBackbone=LLaDA-8B-Instruct, Parallel Decode=Confidence, Block=|B0| = 322026.04 | 39.6 | 20.7 | 99.1 | |
| ConfidenceBackbone=LLaDA-1.5, Parallel Decode=Confidence, Block=|B0| = 322026.04 | 39.4 | 18.9 | 95.2 | |
| VanillaBackbone=LLaDA-8B-Instruct, Parallel Decode=Vanilla, Block=|B0| = 322026.04 | 39.2 | 7.9 | 256 | |
| CAP-DecodingBackbone=LLaDA-8B-Instruct, Parallel Decode=CAP-Decoding, Block=|B0| = 322026.04 | 39.2 | 22.8 | 90.9 | |
| AdaBlockBackbone=LLaDA-1.5, Parallel Decode=Confidence, Block=AdaBlock2026.04 | 38.8 | 18.5 | 97.6 | |
| ConfidenceBackbone=LLaDA-1.5, Parallel Decode=Confidence, Block=|B0| = 642026.04 | 38 | 19.3 | 91.7 | |
| ConfidenceBackbone=LLaDA-8B-Instruct, Parallel Decode=Confidence, Block=|B0| = 642026.04 | 37.8 | 21.1 | 97 | |
| BF16Model=Qwen3-30B-A3B, Quantization=BF162026.04 | 32.2 | — | — | |
| CodeQuantModel=Qwen3-30B-A3B, Quantization Setting=A4W4 Embedding-wise2026.04 | 24.1 | — | — | |
| QuaRotModel=Qwen3-30B-A3B, Quantization Setting=A4W4 Embedding-wise2026.04 | 12.8 | — | — | |
| BF16Model=DeepSeek-V2-Lite, Quantization=BF162026.04 | 12.1 | — | — | |
| CodeQuantModel=DeepSeek-V2-Lite, Quantization Setting=A4W4 Embedding-wise2026.04 | 10.8 | — | — | |
| QuaRotModel=DeepSeek-V2-Lite, Quantization Setting=A4W4 Embedding-wise2026.04 | 9.3 | — | — |