Response correctness and completeness evaluation on Wikipedia
68F1 ScoreHALT
Evaluation results
| Method | Links | |
|---|---|---|
| HALTLLM=Llama-3-70B, n_desired=n_capable2025.06 | 68 | |
| HALTLLM=Mistral-7B, n_desired=n_capable2025.06 | 68 | |
| HALTLLM=Llama-3-8B, n_desired=n_capable2025.06 | 63 | |
| HALTLLM=Gemma-2-9B, n_desired=n_capable2025.06 | 57 | |
| Abstain (IDK)LLM=Llama-3-70B, n_desired=n_capable2025.06 | 44 | |
| FactTuneLLM=Mistral-7B, n_desired=n_capable2025.06 | 44 | |
| FactTuneLLM=Llama-3-8B, n_desired=n_capable2025.06 | 38 | |
| UnchangedLLM=Llama-3-70B, n_desired=n_capable2025.06 | 35 | |
| RandomTrimLLM=Mistral-7B, n_desired=n_capable2025.06 | 34 | |
| HALTLLM=Llama-3-70B, n_desired=n_all2025.06 | 33 | |
| FactTuneLLM=Gemma-2-9B, n_desired=n_capable2025.06 | 31 | |
| HALTLLM=Llama-3-8B, n_desired=n_all2025.06 | 28 | |
| Abstain (IDK)LLM=Llama-3-70B, n_desired=n_all2025.06 | 28 | |
| UnchangedLLM=Llama-3-8B, n_desired=n_capable2025.06 | 26 | |
| UnchangedLLM=Llama-3-70B, n_desired=n_all2025.06 | 26 | |
| RandomTrimLLM=Llama-3-70B, n_desired=n_capable2025.06 | 26 | |
| Abstain (IDK)LLM=Llama-3-8B, n_desired=n_capable2025.06 | 25 | |
| FactTuneLLM=Llama-3-8B, n_desired=n_all2025.06 | 24 | |
| UnchangedLLM=Llama-3-8B, n_desired=n_all2025.06 | 23 | |
| RandomTrimLLM=Llama-3-8B, n_desired=n_capable2025.06 | 21 | |
| UnchangedLLM=Gemma-2-9B, n_desired=n_capable2025.06 | 21 | |
| FactTuneLLM=Mistral-7B, n_desired=n_all2025.06 | 20 | |
| UnchangedLLM=Mistral-7B, n_desired=n_capable2025.06 | 20 | |
| RandomTrimLLM=Gemma-2-9B, n_desired=n_capable2025.06 | 19 | |
| Abstain (IDK)LLM=Gemma-2-9B, n_desired=n_capable2025.06 | 19 | |
| Abstain (IDK)LLM=Mistral-7B, n_desired=n_capable2025.06 | 18 | |
| UnchangedLLM=Gemma-2-9B, n_desired=n_all2025.06 | 15 | |
| Abstain (IDK)LLM=Llama-3-8B, n_desired=n_all2025.06 | 14 | |
| UnchangedLLM=Mistral-7B, n_desired=n_all2025.06 | 14 | |
| HALTLLM=Mistral-7B, n_desired=n_all2025.06 | 13 | |
| Abstain (IDK)LLM=Gemma-2-9B, n_desired=n_all2025.06 | 13 | |
| RandomTrimLLM=Llama-3-70B, n_desired=n_all2025.06 | 12 | |
| Abstain (IDK)LLM=Mistral-7B, n_desired=n_all2025.06 | 11 | |
| HALTLLM=Gemma-2-9B, n_desired=n_all2025.06 | 9 | |
| FactTuneLLM=Gemma-2-9B, n_desired=n_all2025.06 | 8 | |
| RandomTrimLLM=Llama-3-8B, n_desired=n_all2025.06 | 5 | |
| RandomTrimLLM=Mistral-7B, n_desired=n_all2025.06 | 5 | |
| RandomTrimLLM=Gemma-2-9B, n_desired=n_all2025.06 | 2 |