Computer Vision Reasoning on CV-Bench
85.7AccuracyGAMSI_S1+S2
Evaluation Results
| Method | Links | |
|---|---|---|
| GAMSI_S1+S2SI Dataset=950K2026.05 | 85.7 | |
| VST-7B-SFTSI Dataset=4M2026.05 | 85.5 | |
| GazeVLM (Ours)Gaze Bias=Enabled, Base Model=Qwen3-VL-4B2026.05 | 85.4 | |
| VST-3B-SFTSI Dataset=4M2026.05 | 84.6 | |
| GazeVLM (w/o gaze bias)Gaze Bias=None, Base Model=Qwen3-VL-4B2026.05 | 84.5 | |
| SenseNova-SI-InternVL3-8BSI Dataset=800K2026.05 | 84 | |
| GAMSI_S1SI Dataset=800K2026.05 | 83.9 | |
| Qwen3-VL-4BTraining Paradigm=Vanilla Open-source, Parameters=4B2026.05 | 83.4 | |
| SenseNova-SI-Qwen3-VL-8BSI Dataset=8M2026.05 | 83.4 | |
| Qwen3-VL-8B-Instruct2026.05 | 82.4 | |
| Qwen3.5-4B-InstTraining Paradigm=Vanilla Open-source, Parameters=4B2026.05 | 81.1 | |
| Cambrian-S-7BSI Dataset=590K2026.05 | 76.9 | |
| Qwen2.5-VL-7B-Instruct2026.05 | 75.3 | |
| Cambrian-S-3BSI Dataset=590K2026.05 | 75.2 | |
| SpaceR-7BSI Dataset=151K2026.05 | 74.9 | |
| Qwen2.5-VL-7BTraining Paradigm=Vanilla Open-source, Parameters=7B2026.05 | 73.7 | |
| SpatialLadder-3BSI Dataset=260K2026.05 | 73.6 | |
| InternVL3-8B2026.05 | 68.4 | |
| InternVL3-2B2026.05 | 68.1 | |
| Qwen3-VL-2B-Instruct2026.05 | 66.3 | |
| GPT-4oTraining Paradigm=Closed-source2026.05 | 66 | |
| Gemini2.5-Flash-LiteTraining Paradigm=Closed-source2026.05 | 64.9 | |
| Qwen2.5-VL-3B-Instruct2026.05 | 63 | |
| Bagel-7B-MoT2026.05 | 56.7 | |
| Spatial-MLLM-4BSI Dataset=135K2026.05 | 54.7 | |
| ViLaSR-7BSI Dataset=81K2026.05 | 43.5 |