Open-ended Text Generation on Story (test)
0.96Diversity (DIV)Typical Sampling
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Typical SamplingModel=GPT2-XL, Decoding Strategy=typical=0.952022.10 | 0.96 | 0.88 | 0.43 | |
| Typical SamplingModel=OPT-13B, Decoding Strategy=typical=0.952022.10 | 0.95 | 0.91 | 0.46 | |
| Nucleus SamplingModel=GPT2-XL, Decoding Strategy=p=0.952022.10 | 0.94 | 0.91 | 0.46 | |
| Nucleus SamplingModel=OPT-13B, Decoding Strategy=p=0.952022.10 | 0.93 | 0.91 | 0.48 | |
| Top-k SamplingModel=OPT-13B, Decoding Strategy=k=502022.10 | 0.91 | 0.9 | 0.51 | |
| Top-k SamplingModel=GPT2-XL, Decoding Strategy=k=502022.10 | 0.91 | 0.87 | 0.51 | |
| Contrastive DecodingModel=OPT-13B, Decoding Strategy=CD2022.10 | 0.89 | 0.94 | 0.62 | |
| Contrastive SearchModel=GPT2-XL, Decoding Strategy=CS2022.10 | 0.88 | 0.78 | 0.48 | |
| Contrastive DecodingModel=GPT2-XL, Decoding Strategy=CD2022.10 | 0.83 | 0.94 | 0.64 | |
| Contrastive SearchModel=OPT-13B, Decoding Strategy=CS2022.10 | 0.81 | 0.78 | 0.47 | |
| Greedy DecodingModel=OPT-13B, Decoding Strategy=max prob2022.10 | 0.02 | 0.05 | 0.51 | |
| Greedy DecodingModel=GPT2-XL, Decoding Strategy=max prob2022.10 | 0.01 | 0.03 | 0.49 |