Needle-in-a-haystack localization on Pinpoint-Bench
55.2AccuracyPixelEyes
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PixelEyesSize=8B2026.06 | 55.2 | 26.64 | 74.83 | |
| PixelEyesSize=4B2026.06 | 54.73 | 26.13 | 76.91 | |
| Qwen-3-VLSize=8B2026.06 | 49.88 | — | — | |
| Qwen-3-VLSize=4B2026.06 | 46.19 | — | — | |
| Mini-o3Size=7B2026.06 | 44.34 | 8.38 | 78.52 | |
| Gemini-3-FlashSize=-2026.06 | 42.26 | — | — | |
| ThymeSize=7B2026.06 | 40.42 | — | — | |
| DeepEyesSize=7B2026.06 | 39.72 | 14.89 | 20.79 | |
| Qwen-2.5-VLSize=7B2026.06 | 39.03 | — | — | |
| Pixel-ReasonerSize=7B2026.06 | 29.79 | 15.56 | 46.88 |