Language Modeling on WritingPrompts
32MAUVEContinuous Perturbation (w/o DB)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Continuous Perturbation (w/o DB)Base Model=Qwen32026.05 | 32 | — | 93 | |
| Continuous Perturbation (w/ DB)Base Model=Qwen32026.05 | 30 | — | 93 | |
| NEFTuneBase Model=GPT-Neo2026.05 | 30 | — | 102 | |
| Continuous Perturbation (w/o DB)Base Model=GPT-Neo2026.05 | 29 | — | 97 | |
| Continuous Perturbation (w/ DB)Base Model=GPT-Neo2026.05 | 29 | — | 97 | |
| NEFTuneBase Model=Qwen32026.05 | 28 | — | 113 | |
| VanillaBase Model=GPT-Neo2026.05 | 24 | — | 101 | |
| Discrete PerturbationBase Model=GPT-Neo2026.05 | 23 | — | 123 | |
| Continuous Perturbation (w/o DB)Base Model=OPT2026.05 | 21 | — | 125 | |
| Discrete PerturbationBase Model=Qwen32026.05 | 21 | — | 148 | |
| Continuous Perturbation (w/ DB)Base Model=OPT2026.05 | 20 | — | 121 | |
| MLE (ours)Model=GPT-22026.05 | 17 | 38 | — | |
| MLE (ours)Model=GPT-Neo-1.3B2026.05 | 17 | 38 | — | |
| VanillaBase Model=Qwen32026.05 | 17 | — | 120 | |
| NEFTuneBase Model=OPT2026.05 | 16 | — | 160 | |
| Discrete PerturbationBase Model=OPT2026.05 | 15 | — | 179 | |
| MLEModel=GPT-Neo-1.3B2026.05 | 14 | 36 | — | |
| MixCEModel=GPT-Neo-1.3B2026.05 | 14 | 37 | — | |
| TVDModel=GPT-Neo-1.3B2026.05 | 14 | 37 | — | |
| MLE (ours)Model=OPT-125M2026.05 | 13 | 38 | — | |
| EMOModel=GPT-22026.05 | 13 | 37 | — | |
| BrierModel=GPT-Neo-1.3B2026.05 | 13 | 36 | — | |
| EMOModel=GPT-Neo-1.3B2026.05 | 13 | 35 | — | |
| VanillaBase Model=OPT2026.05 | 13 | — | 164 | |
| MLEModel=OPT-125M2026.05 | 11 | 36 | — | |
| BrierModel=OPT-125M2026.05 | 11 | 36 | — | |
| BrierModel=GPT-22026.05 | 11 | 36 | — | |
| MixCEModel=OPT-125M2026.05 | 10 | 36 | — | |
| TVDModel=OPT-125M2026.05 | 10 | 36 | — | |
| EMOModel=OPT-125M2026.05 | 10 | 36 | — | |
| MLEModel=GPT-22026.05 | 10 | 36 | — | |
| MixCEModel=GPT-22026.05 | 10 | 36 | — | |
| TVDModel=GPT-22026.05 | 9 | 36 | — |