Negative Retrieval on NegBench
64.62Image-to-Text Recall@1HANCLIP (SmartCLIP backbone)
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| HANCLIP (SmartCLIP backbone)ViT=L/14, Setting=+ Ours2026.06 | 64.62 | 85.16 | 91.22 | 46.39 | 71.1 | 80.02 | 438.51 | |
| SmartCLIPViT=L/14, Setting=Base2026.06 | 63 | 84 | 90.02 | 43.75 | 68.77 | 77.96 | 427.49 | |
| HANCLIP (HiMo-CLIP backbone)ViT=L/14, Setting=+ Ours2026.06 | 62.7 | 84.36 | 90.72 | 45.36 | 70.63 | 79.31 | 433.08 | |
| NegationCLIPViT=L/14, Setting=Base2026.06 | 62.58 | 85.7 | 91.66 | 45.08 | 70.66 | 80.38 | 436.06 | |
| HiMo-CLIPViT=L/14, Setting=Base2026.06 | 60.96 | 83.12 | 89.92 | 43.18 | 68.35 | 77.28 | 422.8 | |
| HANCLIP (LongCLIP backbone)ViT=L/14, Setting=+ Ours2026.06 | 60.9 | 83.58 | 89.74 | 43.79 | 68.66 | 77.86 | 424.54 | |
| NegationCLIPViT=B/16, Setting=Base2026.06 | 59.52 | 83.22 | 89.62 | 41.67 | 68.49 | 78.38 | 420.89 | |
| HANCLIP (SmartCLIP backbone)ViT=B/16, Setting=+ Ours2026.06 | 58.82 | 81.1 | 88.62 | 41.18 | 66.99 | 76.85 | 413.56 | |
| NegCLIPCC12MViT=B/32, Setting=Base2026.06 | 58.34 | 82.04 | 88.96 | 41.18 | 68.1 | 78.26 | 416.88 | |
| SmartCLIPViT=B/16, Setting=Base2026.06 | 57.44 | 80.42 | 87.82 | 38.48 | 64.56 | 74.44 | 403.17 | |
| LongCLIPViT=L/14, Setting=Base2026.06 | 56.36 | 79.28 | 86.8 | 40.95 | 66.1 | 75.67 | 405.15 | |
| NegationCLIPViT=B/32, Setting=Base2026.06 | 56.06 | 80.72 | 88.24 | 38.59 | 65.8 | 76.39 | 405.8 | |
| NegCLIPViT=B/32, Setting=Base2026.06 | 56.04 | 80.52 | 88.18 | 39.88 | 67.17 | 77.34 | 409.14 | |
| HANCLIP (LongCLIP backbone)ViT=B/16, Setting=+ Ours2026.06 | 55.56 | 78.78 | 86.2 | 37.91 | 63.66 | 73.65 | 395.77 | |
| HANCLIP (CLIP backbone)ViT=L/14, Setting=+ Ours2026.06 | 53.4 | 76.76 | 84.72 | 32.67 | 56.98 | 67.27 | 371.8 | |
| LongCLIPViT=B/16, Setting=Base2026.06 | 52.08 | 76.28 | 83.96 | 35.83 | 61.37 | 71.7 | 381.22 | |
| CoNCLIPViT=L/14, Setting=Base2026.06 | 51.94 | 75.24 | 82.58 | 33.09 | 57.13 | 67.87 | 367.85 | |
| CLIPViT=L/14, Setting=Base2026.06 | 50.88 | 74.26 | 82.42 | 29.06 | 52.38 | 62.95 | 351.96 | |
| HANCLIP (CLIP backbone)ViT=B/16, Setting=+ Ours2026.06 | 50.56 | 75.2 | 83.64 | 29.73 | 53.73 | 64.7 | 357.58 | |
| CLIPCC12MViT=B/32, Setting=Base2026.06 | 48.72 | 73.24 | 81.82 | 30.84 | 55.81 | 66.63 | 357.06 | |
| HANCLIP (CLIP backbone)ViT=B/32, Setting=+ Ours2026.06 | 48.53 | 73.34 | 82.02 | 28.83 | 53.34 | 64.55 | 350.65 | |
| CLIPViT=B/16, Setting=Base2026.06 | 47.86 | 72.36 | 81.46 | 27.05 | 50.26 | 61.31 | 340.3 | |
| CE_CLIPViT=B/32, Setting=Base2026.06 | 47.26 | 76.1 | 85.64 | 39.21 | 66.39 | 76.65 | 391.24 | |
| CoNCLIPViT=B/32, Setting=Base2026.06 | 45.54 | 69.22 | 77.54 | 26.76 | 51.22 | 62.56 | 332.83 | |
| CLIPViT=B/32, Setting=Base2026.06 | 45.34 | 69.72 | 79.08 | 24.93 | 47.93 | 59.41 | 326.42 |