Question Answering on TYDIQAGP
42.85Factual AccuracyARREST-HB
Evaluation Results
| Method | Links | |
|---|---|---|
| ARREST-HBLLM=Vicuna 7B2026.01 | 42.85 | |
| DOLALLM=Vicuna 7B2026.01 | 28.8 | |
| Activation DecodingLLM=Vicuna 7B2026.01 | 27.12 | |
| ARREST-HBLLM=LLaMa-3.1 8B2026.01 | 22.42 | |
| ITILLM=Vicuna 7B2026.01 | 19.5 | |
| Self-ReflectionLLM=Vicuna 7B2026.01 | 19.27 | |
| COVELLM=Vicuna 7B2026.01 | 19.1 | |
| Self-ReflectionLLM=LLaMa-3.1 8B2026.01 | 18.69 | |
| ARREST-HBLLM=LLaMa-2 7B2026.01 | 17.8 | |
| DOLALLM=LLaMa-2 7B2026.01 | 16.92 | |
| Base ModelLLM=Vicuna 7B2026.01 | 16.39 | |
| ITILLM=LLaMa-2 7B2026.01 | 15.96 | |
| COVELLM=LLaMa-3.1 8B2026.01 | 13.64 | |
| Activation DecodingLLM=LLaMa-3.1 8B2026.01 | 13.64 | |
| ITILLM=LLaMa-3.1 8B2026.01 | 13.04 | |
| DOLALLM=LLaMa-3.1 8B2026.01 | 10.96 | |
| Base ModelLLM=LLaMa-3.1 8B2026.01 | 10.49 | |
| Self-ReflectionLLM=LLaMa-2 7B2026.01 | 10.2 | |
| COVELLM=LLaMa-2 7B2026.01 | 9.53 | |
| Activation DecodingLLM=LLaMa-2 7B2026.01 | 6.92 | |
| Base ModelLLM=LLaMa-2 7B2026.01 | 3.5 |