Concepts
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Text-to-Image Personalization | Concepts dataset | CLIP-I Score0.717 | 14 | |
| Steering Detection | 121 concepts out-of-distribution suites (held-out) | Detection Rate95.5 | 14 | |
| Text-to-image concept steering | 167 held-out concepts (test) | Stage 1 Input Fidelity (VLM)83.1 | 4 | |
| Concept Extraction | 17 concepts (evaluation set) | Model Preference Count (Handoff)11 | 2 |
Showing 4 of 4 rows