Long-form generation on Bio (PIA and Poison Metrics)
69.8PIA RLLMJ ScoreDecoding_c
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Decoding_cLLM=Mistral-7B2024.05 | 69.8 | 0 | 71 | 0 | |
| Decoding_cLLM=Llama2-7B2024.05 | 67.6 | 0 | 67.8 | 0 | |
| KeywordLLM=Mistral-7B2024.05 | 64.8 | 0 | 61.6 | 0 | |
| KeywordLLM=Llama2-7B2024.05 | 59.2 | 0 | 63.4 | 0 | |
| VanillaLLM=Mistral-7B2024.05 | 29 | 100 | 56 | 86 | |
| VanillaLLM=Llama2-7B2024.05 | 18.2 | 98 | 42.4 | 44 |