Label Error Detection on AlleNoise 1,000-item balanced subset
100PrecisionAdjudicator
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| AdjudicatorKnowledge Graph=Full2025.12 | 100 | 98 | 99 | |
| AdjudicatorKnowledge Graph=None2025.12 | 85 | 45 | 59 | |
| Single LLM2025.12 | 66 | 38 | 48 |
| Method | Links | |||
|---|---|---|---|---|
| AdjudicatorKnowledge Graph=Full2025.12 | 100 | 98 | 99 | |
| AdjudicatorKnowledge Graph=None2025.12 | 85 | 45 | 59 | |
| Single LLM2025.12 | 66 | 38 | 48 |