Specification Revision Conformance on combined real-LLM corpus
-10Mean Difference (Delta)Absorber
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| AbsorberComparison Baseline (vs.)=Full-Rest., Evaluated Attribute=waste2026.04 | -10 | -2.05 | 10.89 | |
| AbsorberComparison Baseline (vs.)=Oracle, Evaluated Attribute=quality2026.04 | -0.75 | -0.77 | 0.95 | |
| AbsorberComparison Baseline (vs.)=Full-Rest., Evaluated Attribute=quality2026.04 | -0.11 | -0.12 | 0.27 | |
| AbsorberComparison Baseline (vs.)=Naive, Evaluated Attribute=quality2026.04 | 0.33 | 0.3 | 0.13 | |
| AbsorberComparison Baseline (vs.)=Oracle, Evaluated Attribute=waste2026.04 | 0.74 | 2.14 | 0.69 | |
| AbsorberComparison Baseline (vs.)=Ignore, Evaluated Attribute=quality2026.04 | 1.38 | 1.32 | 1.19 |