Natural Language Explanation Generation on HealthVer (test)
0.033FaithfulnessCLUE-Span+Steering
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| CLUE-Span+SteeringModel=Qwen2.5-14B2025.05 | 0.033 | 0.44 | 0.53 | 0.8 | |
| CLUE-Span+SteeringModel=Gemma-2-9B-IT2025.05 | 0.021 | 0.39 | 0.5 | 0.85 | |
| CLUE-Span+SteeringModel=OLMo-2-1124-13B2025.05 | 0.02 | 0.23 | 0.77 | 0.68 | |
| CLUE-SpanModel=Gemma-2-9B-IT2025.05 | 0.007 | 0.34 | 0.59 | 0.82 | |
| CLUE-SpanModel=Qwen2.5-14B2025.05 | 0.006 | 0.33 | 0.68 | 0.75 | |
| CLUE-SpanModel=OLMo-2-1124-13B2025.05 | 0.005 | 0.1 | 0.83 | 0.61 | |
| PromptBaselineModel=Qwen2.5-14B2025.05 | -0.028 | — | — | 0.74 | |
| PromptBaselineModel=OLMo-2-1124-13B2025.05 | -0.1 | — | — | 0.55 | |
| PromptBaselineModel=Gemma-2-9B-IT2025.05 | -0.105 | — | — | 0.66 |