Error Localization on Controlled Injected Benchmark 1266 cases
76.7Top-1 Major Category HitOptArgus
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OptArgusArchitecture=Taxonomy-grounded multi-agent2026.05 | 76.7 | 47.3 | 40.3 | 1.224 | |
| Single-Agent DetectorArchitecture=Single-agent monolithic2026.05 | 72.3 | 43.1 | 33.9 | 2.151 |