Entity Binding in Tool Use on 60-task entity binding diagnostic testbed 1.0 (aggregate)
75Task Success RateSemantic filter
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Semantic filter2026.06 | 75 | 75.7 | 0 | 24 | 1 | 0 | 1.037 | |
| CMTF only2026.06 | 74.3 | 74.3 | 0 | 25.7 | 0 | 0 | 1.11 | |
| Direct2026.06 | 74 | 74 | 0 | 26 | 0 | 0 | 1.123 | |
| Entity retrieval2026.06 | 74 | 74 | 0 | 26 | 0 | 0 | 1.123 | |
| Confidence gateEntity-aware=true2026.06 | 31.7 | 40 | 0 | 0 | 68.3 | 0 | 0 | |
| Entity CMTF+Prov.Provenance=true, Entity-aware=true2026.06 | 26 | 34.3 | 0 | 0 | 74 | 0 | 0 |