Code Generation on LiveCodeBench (Speedup and Mean Acceptance Length)
10.28Mean Acceptance Length (τ)DFlash+DDTree
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DFlash+DDTreeModel=Qwen3-8B, Temperature=0.0, Node Budget=Best from {16, 32, 64, 128, 256, 512, 1024}2026.04 | 10.28 | 7.1 | |
| DFlash+DDTreeModel=Qwen3-4B, Temperature=0.0, Node Budget=Best from {16, 32, 64, 128, 256, 512, 1024}2026.04 | 9.79 | 6.78 | |
| DFlash+DDTreeModel=Qwen3-8B, Temperature=1.0, Node Budget=Best from {16, 32, 64, 128, 256, 512, 1024}2026.04 | 9.53 | 6.46 | |
| DFlash+DDTreeModel=Qwen3-4B, Temperature=1.0, Node Budget=Best from {16, 32, 64, 128, 256, 512, 1024}2026.04 | 9.44 | 6.44 | |
| DFlash+DDTreeModel=Qwen3-Coder-30B-A3B-Instruct, Temperature=0.0, Node Budget=Best from {16, 32, 64, 128, 256, 512, 1024}2026.04 | 8.63 | 6.54 | |
| DFlash+DDTreeModel=Qwen3-Coder-30B-A3B-Instruct, Temperature=1.0, Node Budget=Best from {16, 32, 64, 128, 256, 512, 1024}2026.04 | 7.79 | 5.47 | |
| DFlashModel=Qwen3-8B, Temperature=0.02026.04 | 7.22 | 5.02 | |
| DFlashModel=Qwen3-4B, Temperature=0.02026.04 | 7.02 | 4.97 | |
| DFlashModel=Qwen3-8B, Temperature=1.02026.04 | 6.61 | 4.52 | |
| DFlashModel=Qwen3-4B, Temperature=1.02026.04 | 6.53 | 4.53 | |
| DFlashModel=Qwen3-Coder-30B-A3B-Instruct, Temperature=0.02026.04 | 6.32 | 4.72 | |
| DFlashModel=Qwen3-Coder-30B-A3B-Instruct, Temperature=1.02026.04 | 5.57 | 3.77 | |
| Qwen3 8B + EAGLE3Drafter Norm=Post, Decoding Mode=no-thinking2026.05 | 4.71 | — | |
| Qwen3 8B + EAGLE3Drafter Norm=Pre, Decoding Mode=no-thinking2026.05 | 4.59 | — | |
| Llama 3.1 8B + EAGLE3Model=Llama 3.1 8B, Decoding Method=SGLang + EAGLE3, Steps=7, Top-k=1, Draft Length=8, Normalization=Post-norm2026.05 | 4.22 | — | |
| Llama 3.1 8B + EAGLE3Model=Llama 3.1 8B, Decoding Method=SGLang + EAGLE3, Steps=7, Top-k=1, Draft Length=8, Normalization=Pre-norm2026.05 | 4.16 | — | |
| GPT-oss 20B + EAGLE3Normalization=Post-norm, Reasoning Effort=low2026.05 | 3.26 | — | |
| Qwen3 8B + EAGLE3Drafter Norm=Post, Decoding Mode=thinking2026.05 | 3.14 | — | |
| GPT-oss 20B + EAGLE3Normalization=Pre-norm, Reasoning Effort=low2026.05 | 3.12 | — | |
| Qwen3 8B + EAGLE3Drafter Norm=Pre, Decoding Mode=thinking2026.05 | 3.06 | — | |
| GPT-oss 20B + EAGLE3Normalization=Post-norm, Reasoning Effort=medium2026.05 | 2.79 | — | |
| GPT-oss 20B + EAGLE3Normalization=Pre-norm, Reasoning Effort=medium2026.05 | 2.69 | — |