Question Answering on TruthfulQA (Factual Accuracy)
47Factual AccuracyARREST-HB
Evaluation results
| Method | Links | |
|---|---|---|
| ARREST-HBLLM=LLaMa-2 7B2026.01 | 47 | |
| DOLALLM=LLaMa-2 7B2026.01 | 45.43 | |
| ITILLM=LLaMa-2 7B2026.01 | 44.31 | |
| Activation DecodingLLM=LLaMa-2 7B2026.01 | 43.53 | |
| ARREST-HBLLM=Vicuna 7B2026.01 | 40.4 | |
| Activation DecodingLLM=Vicuna 7B2026.01 | 34.57 | |
| Self-ReflectionLLM=Vicuna 7B2026.01 | 33.34 | |
| ITILLM=Vicuna 7B2026.01 | 31.9 | |
| COVELLM=Vicuna 7B2026.01 | 30.28 | |
| Self-ReflectionLLM=LLaMa-2 7B2026.01 | 28.16 | |
| Self-ReflectionLLM=LLaMa-3.1 8B2026.01 | 27.84 | |
| COVELLM=LLaMa-3.1 8B2026.01 | 27.1 | |
| ARREST-HBLLM=LLaMa-3.1 8B2026.01 | 26.93 | |
| DOLALLM=LLaMa-3.1 8B2026.01 | 25.3 | |
| DOLALLM=Vicuna 7B2026.01 | 24.63 | |
| Base ModelLLM=Vicuna 7B2026.01 | 24.11 | |
| Activation DecodingLLM=LLaMa-3.1 8B2026.01 | 23.84 | |
| ITILLM=LLaMa-3.1 8B2026.01 | 23.5 | |
| Base ModelLLM=LLaMa-3.1 8B2026.01 | 21.04 | |
| COVELLM=LLaMa-2 7B2026.01 | 15 | |
| Base ModelLLM=LLaMa-2 7B2026.01 | 13.83 |