List your company

Natural Language Explanation Generation on DRUID (test)

0.102FaithfulnessCLUE-Span+Steering
-0.13928-0.07664-0.0140.04864May 23, 2025

Evaluation results

MethodLinks
CLUE-Span+SteeringModel=Qwen2.5-14B2025.050.10228200.77
CLUE-Span+SteeringModel=OLMo-2-1124-13B2025.050.09915700.69
CLUE-Span+SteeringModel=Gemma-2-9B-IT2025.050.09830470.81
CLUE-SpanModel=Qwen2.5-14B2025.050.08920380.78
CLUE-SpanModel=Gemma-2-9B-IT2025.050.04323430.76
CLUE-SpanModel=OLMo-2-1124-13B2025.050.0148790.65
PromptBaselineModel=Qwen2.5-14B2025.05-0.08——0.6
PromptBaselineModel=Gemma-2-9B-IT2025.05-0.12——0.57
PromptBaselineModel=OLMo-2-1124-13B2025.05-0.13——0.53