Infilling on ROCStories
25.4ROUGE-1DiffuLLaMA + P2-self
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| DiffuLLaMA + P2-selfModel=DiffuLLaMA, Parameters=7B, Sampling Strategy=P2-self2025.02 | 25.4 | 7.1 | 23.4 | — | — | |
| DiffuLlamaSize=7B, Type=DD2026.01 | 23.3 | 5.5 | 21.2 | — | — | |
| MDLMSize=130M, Tokens=157B2026.07 | 22.2 | 2.3 | 15.2 | — | 72.6 | |
| DiffuLLaMAModel=DiffuLLaMA, Parameters=7B, Sampling Strategy=ancestral sampling2025.02 | 20.3 | 2.8 | 18.2 | — | — | |
| A3Size=8B, Type=A32026.01 | 19.2 | 4.6 | 18.6 | — | — | |
| DiffuGPT-MModel=DiffuGPT-M, Parameters=355M2025.02 | 18.7 | 2.7 | 17 | — | — | |
| AR TransformerSize=130M, Tokens=157B2026.07 | 18.3 | 1.3 | 13.1 | — | 158.4 | |
| SW-SetDLM S ≤ 32Size=110M, Tokens=157B2026.07 | 18.1 | 1.3 | 13.2 | — | 123.1 | |
| ASSDSize=110M, Tokens=45B2026.07 | 18 | 1.4 | 13.2 | — | — | |
| DiffuGPT-SSize=127M, Tokens=n/a2026.07 | 16.4 | 2 | 14.2 | — | — | |
| SEDD-SSize=170M, Tokens=210B2026.07 | 16.2 | 1.3 | 12.2 | — | — | |
| BD3LM S = 16Size=110M, Tokens=157B2026.07 | 15.8 | 0.8 | 11.1 | — | 114.2 | |
| DiffuGPT-SModel=DiffuGPT-S, Parameters=127M2025.02 | 13.7 | 1.4 | 12.6 | — | — | |
| GPT2-SSize=127M, Tokens=n/a2026.07 | 13.5 | 0.6 | 10.2 | — | — | |
| SEDD-MModel=SEDD-M, Parameters=424M2025.02 | 13.1 | 1.4 | 12.2 | — | — | |
| PlaidSize=1B, Type=CD2026.01 | 12.1 | 1.1 | 11.2 | — | — | |
| Plaid1BModel=Plaid1B, Parameters=1.3B2025.02 | 12.1 | 1.1 | 11.2 | — | — | |
| SEDD-SModel=SEDD-S, Parameters=170M2025.02 | 11.9 | 0.7 | 10.9 | — | — | |
| A3Size=1B, Type=A32026.01 | 11.8 | 1.7 | 11.1 | — | — | |
| Llama-3.1Size=8B, Type=AR2026.01 | 11.7 | 2.3 | 10.5 | — | — | |
| DreamSize=7B, Type=DD2026.01 | 11.7 | 2.3 | 10.5 | — | — | |
| LLaMA2Model=LLaMA2, Parameters=7B2025.02 | 11.6 | 2.1 | 10.5 | — | — | |
| A3Size=3B, Type=A32026.01 | 11.3 | 2.3 | 10.2 | — | — | |
| GPT2-MModel=GPT2-M, Parameters=355M2025.02 | 8.6 | 0.9 | 8.2 | — | — | |
| GPT2-SModel=GPT2-S, Parameters=127M2025.02 | 7.8 | 0.8 | 7.4 | — | — | |
| BaseCPT=Base, SFT=None2026.05 | — | — | — | 6.94 | — | |
| BaseCPT=Base, SFT=Base2026.05 | — | — | — | 9.87 | — | |
| Base-GaussianCPT=Base, SFT=Gaussian2026.05 | — | — | — | 9.48 | — | |
| GaussianCPT=Gaussian, SFT=None2026.05 | — | — | — | 7.31 | — | |
| Gaussian-BaseCPT=Gaussian, SFT=Base2026.05 | — | — | — | 12.18 | — | |
| Gaussian-GaussianCPT=Gaussian, SFT=Gaussian2026.05 | — | — | — | 12.62 | — |