Code Generation on HumanEval 0-shot (test)
45.7AccuracyPrefix-Cache
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Prefix-CacheGen Length=512, Evaluation Protocol=0-shot2026.01 | 45.7 | — | — | 21.4 | 1.8 | — | — | — | — | — | |
| Fast-dLLM (threshold)Generation Length=5122025.12 | 45.12 | — | — | — | — | 94.41 | 1 | — | — | — | |
| dKV-CacheGen Length=512, Evaluation Protocol=0-shot2026.01 | 45.1 | — | — | 32.9 | 1.2 | — | — | — | — | — | |
| Fast-dLLMGen Length=512, Evaluation Protocol=0-shot2026.01 | 44.5 | — | — | 9 | 4.4 | — | — | — | — | — | |
| Streaming-dLLMGen Length=512, Evaluation Protocol=0-shot2026.01 | 44.5 | — | — | 5.2 | 7.6 | — | — | — | — | — | |
| LLaDAGen Length=512, Evaluation Protocol=0-shot2026.01 | 43.9 | — | — | 39.4 | 1 | — | — | — | — | — | |
| Credit DecodingBackbone=LLaDA-8B-Instruct, KV Cache=true, Generation Length=2562026.05 | 43.7 | — | — | — | 3.8 | 24.7 | — | — | — | — | |
| Streaming-dLLMGen Length=256, Evaluation Protocol=0-shot2026.01 | 43.4 | — | — | 2.5 | 5.3 | — | — | — | — | — | |
| Fast-dLLMGen Length=256, Evaluation Protocol=0-shot2026.01 | 43.3 | — | — | 3.4 | 3.9 | — | — | — | — | — | |
| Fast-dLLMBackbone=LLaDA-8B-Instruct, KV Cache=true, Generation Length=2562026.05 | 43.3 | — | — | — | 3.7 | 24.1 | — | — | — | — | |
| CadLLMGeneration Length=5122025.12 | 43.29 | — | — | — | — | 163.72 | 1.74 | — | — | — | |
| TSPDBackbone=LLaDA-8B-Instruct, KV Cache=true, Generation Length=2562026.05 | 43 | — | — | — | 8.2 | 53.3 | — | — | — | — | |
| Prefix-CacheGen Length=256, Evaluation Protocol=0-shot2026.01 | 42.7 | — | — | 7.9 | 1.7 | — | — | — | — | — | |
| TSPDBackbone=LLaDA-8B-Instruct, KV Cache=false, Generation Length=2562026.05 | 42.5 | — | — | — | 4.2 | 27.3 | — | — | — | — | |
| LLaDAGen Length=256, Evaluation Protocol=0-shot2026.01 | 41.5 | — | — | 13.3 | 1 | — | — | — | — | — | |
| VanillaBackbone=LLaDA-8B-Instruct, KV Cache=true, Generation Length=2562026.05 | 41.5 | — | — | — | 1 | 6.5 | — | — | — | — | |
| ProphetBackbone=LLaDA-8B-Instruct, KV Cache=false, Generation Length=2562026.05 | 41.5 | — | — | — | 1.2 | 7.8 | — | — | — | — | |
| Fast-dLLM (factor)Generation Length=5122025.12 | 41.46 | — | — | — | — | 131.14 | 1.38 | — | — | — | |
| Learn2PDBackbone=LLaDA-8B-Instruct, KV Cache=false, Generation Length=2562026.05 | 40.3 | — | — | — | 3.6 | 23.4 | — | — | — | — | |
| dKV-CacheGen Length=256, Evaluation Protocol=0-shot2026.01 | 39.6 | — | — | 12.9 | 1 | — | — | — | — | — | |
| Fast-dLLM (threshold)Generation Length=2562025.12 | 37.19 | — | — | — | — | 96.84 | 1 | — | — | — | |
| CadLLMGeneration Length=2562025.12 | 35.97 | — | — | — | — | 220.81 | 2.28 | — | — | — | |
| Fast-dLLM (factor)Generation Length=2562025.12 | 32.92 | — | — | — | — | 132.28 | 1.37 | — | — | — | |
| CDLM-LLaDAShots=0-shot2025.11 | — | — | — | 1.9 | — | 50.9 | — | 32.3 | 95.1 | 40.2 | |
| d2CacheModel=LLaDA-Inst, Few-shot=0-shot, Gen. Len.=512, Decoding Strategy=Semi-AR parallel decoding2025.09 | — | 46.6 | — | 11.89 | 9.7 | 48.39 | — | — | — | — | |
| d2CacheModel=Dream-Inst, Few-shot=0-shot, Gen. Len.=512, Decoding Strategy=Semi-AR parallel decoding2025.09 | — | 57.3 | — | 6.3 | 23.8 | 104.4 | — | — | — | — | |
| diffu-GRPO(d1)Sequence Length=128, Few-shot=02025.12 | — | 25.6 | — | — | — | — | — | — | — | — | |
| diffu-GRPO(d1)Sequence Length=256, Few-shot=02025.12 | — | 36 | — | — | — | — | — | — | — | — | |
| diffu-GRPO(d1)Sequence Length=512, Few-shot=02025.12 | — | 50 | — | — | — | — | — | — | — | — | |
| diffu-GRPO(d1)Few-shot=02025.12 | — | — | 37.2 | — | — | — | — | — | — | — | |
| dLLM-CacheShots=0-shot, cache_mode=dLLM-Cache2025.11 | — | — | — | 10.8 | — | 12.6 | — | 256 | 135.9 | 39 | |
| DualCacheBackbone=LLaDA-8B-Base, Speedup=7.4×2026.03 | — | 32.32 | — | — | — | 177.4 | — | — | — | — | |
| ES-dLLMBackbone=LLaDA-8B-Base, Speedup=9.8×2026.03 | — | 31.71 | — | — | — | 232.66 | — | — | — | — | |
| ESPOSequence Length=128, Few-shot=02025.12 | — | 28.1 | — | — | — | — | — | — | — | — | |
| ESPOSequence Length=256, Few-shot=02025.12 | — | 42.1 | — | — | — | — | — | — | — | — | |
| ESPOSequence Length=512, Few-shot=02025.12 | — | 50 | — | — | — | — | — | — | — | — | |
| ESPOFew-shot=02025.12 | — | — | 40.1 | — | — | — | — | — | — | — | |
| Fast-dLLM (Par.)Shots=0-shot, parallel decoding=true2025.11 | — | — | — | 4.6 | — | 17.9 | — | 100.3 | 83.1 | 36.6 | |
| Fast-dLLM (Par.+D.C.)Shots=0-shot, parallel decoding=true, dual-cache KV=true2025.11 | — | — | — | 4.2 | — | 18.9 | — | 97.7 | 80.2 | 36 | |
| LLaDASequence Length=128, Few-shot=02025.12 | — | 26.8 | — | — | — | — | — | — | — | — | |
| LLaDASequence Length=256, Few-shot=02025.12 | — | 37.8 | — | — | — | — | — | — | — | — | |
| LLaDASequence Length=512, Few-shot=02025.12 | — | 48.8 | — | — | — | — | — | — | — | — | |
| LLaDAFew-shot=02025.12 | — | — | 37.8 | — | — | — | — | — | — | — | |
| LLaDABackbone=LLaDA-8B-Base, Speedup=1.0×2026.03 | — | 32.32 | — | — | — | 23.84 | — | — | — | — | |
| LLaDA-1.5Sequence Length=128, Few-shot=02025.12 | — | 29.3 | — | — | — | — | — | — | — | — | |
| LLaDA-1.5Sequence Length=256, Few-shot=02025.12 | — | 39.6 | — | — | — | — | — | — | — | — | |
| LLaDA-1.5Sequence Length=512, Few-shot=02025.12 | — | 51.9 | — | — | — | — | — | — | — | — | |
| LLaDA-1.5Few-shot=02025.12 | — | — | 40.3 | — | — | — | — | — | — | — | |
| LLaDA-8B-InstructShots=0-shot2025.11 | — | — | — | 11.3 | — | 7.4 | — | 256 | 83.9 | 37.8 | |
| wd1Sequence Length=128, Few-shot=02025.12 | — | 34.7 | — | — | — | — | — | — | — | — | |
| wd1Sequence Length=256, Few-shot=02025.12 | — | 38.4 | — | — | — | — | — | — | — | — | |
| wd1Sequence Length=512, Few-shot=02025.12 | — | 38.4 | — | — | — | — | — | — | — | — | |
| wd1Few-shot=02025.12 | — | — | 37.2 | — | — | — | — | — | — | — |