Object Center Localization on PixMo Points (test)
6.3Median DistanceQwen2.5-VL-7B-Instr. (NTP + LLL)
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen2.5-VL-7B-Instr. (NTP + LLL)Backbone=Qwen2.5-VL-7B-Instr., Training=Fine-tuned on RefCOCO (NTP + LLL), Evaluation=Zero-shot2026.02 | 6.3 | |
| Qwen2.5-VL-7B-Instr. (NTP)Backbone=Qwen2.5-VL-7B-Instr., Training=Fine-tuned on RefCOCO (NTP), Evaluation=Zero-shot2026.02 | 6.95 | |
| Qwen2.5-VL-7B-Instr. (Base Model)Backbone=Qwen2.5-VL-7B-Instr., Training=Zero-shot Base, Evaluation=Zero-shot2026.02 | 7.07 | |
| GPT-4oEvaluation=Zero-shot2026.02 | 7.07 | |
| Molmo-7B-DEvaluation=Zero-shot2026.02 | 7.13 | |
| Gemini-2.0 FlashEvaluation=Zero-shot2026.02 | 7.21 | |
| LLaVA-v1.5-7B (NTP + LLL)Backbone=LLaVA-v1.5-7B, Training=Fine-tuned on RefCOCO (NTP + LLL), Evaluation=Zero-shot2026.02 | 9.21 | |
| LLaVA-v1.5-7B (NTP)Backbone=LLaVA-v1.5-7B, Training=Fine-tuned on RefCOCO (NTP), Evaluation=Zero-shot2026.02 | 12.95 | |
| LLaVA-v1.5-7B (Base Model)Backbone=LLaVA-v1.5-7B, Training=Zero-shot Base, Evaluation=Zero-shot2026.02 | 13.6 |