List your company

Question Answering on COQA (factual accuracy)

28.27Factual AccuracyARREST-HB
12.6716.7220.7724.82Jan 7, 2026

Evaluation results

MethodLinks
ARREST-HBLLM=Vicuna 7B2026.0128.27
DOLALLM=Vicuna 7B2026.0127.5
Activation DecodingLLM=Vicuna 7B2026.0125.02
Self-ReflectionLLM=Vicuna 7B2026.0123.69
ARREST-HBLLM=LLaMa-2 7B2026.0121.59
ITILLM=LLaMa-2 7B2026.0120.37
ARREST-HBLLM=LLaMa-3.1 8B2026.0119.71
DOLALLM=LLaMa-2 7B2026.0118.3
DOLALLM=LLaMa-3.1 8B2026.0118.3
ITILLM=LLaMa-3.1 8B2026.0116.54
Activation DecodingLLM=LLaMa-2 7B2026.0116.3
Activation DecodingLLM=LLaMa-3.1 8B2026.0116.3
Self-ReflectionLLM=LLaMa-2 7B2026.0116
COVELLM=LLaMa-2 7B2026.0115.5
Base ModelLLM=LLaMa-2 7B2026.0115.1
COVELLM=Vicuna 7B2026.0115.04
COVELLM=LLaMa-3.1 8B2026.0115
ITILLM=Vicuna 7B2026.0114.45
Base ModelLLM=Vicuna 7B2026.0114.4
Self-ReflectionLLM=LLaMa-3.1 8B2026.0113.69
Base ModelLLM=LLaMa-3.1 8B2026.0113.27