Multi-image Grounding on MIG-Bench 1.0 (full)
99.5Static ScoreHuman
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Human2026.04 | 99.5 | 97.87 | 98 | 100 | 96.88 | 100 | 98.99 | 91.06 | 92.08 | 97.44 | 97.18 | |
| Migician-7BBase MLLM=Qwen2-VL2026.04 | 65.15 | 46.81 | 84.19 | 70.73 | 60.07 | 74.31 | 76.77 | 66.53 | 59.41 | 34.19 | 63.82 | |
| CGC-8BBase MLLM=Qwen3-VL2026.04 | 55.3 | 54.26 | 78.31 | 78.18 | 66.32 | 70.16 | 80.81 | 76.35 | 73.27 | 42.74 | 67.57 | |
| CGC-7BBase MLLM=Qwen2-VL2026.04 | 48.86 | 48.94 | 77.21 | 51.45 | 59.72 | 58.02 | 61.62 | 71.94 | 56.44 | 36.75 | 55.5 | |
| Qwen2-VL-72BModel Scale=Large-Scale Models (≥70B)2026.04 | 46.12 | 46.81 | 64.46 | 26.73 | 22.57 | 18.62 | 33.33 | 62.53 | 50.5 | 17.09 | 38.88 | |
| Qwen2.5-VL-72BModel Scale=Large-Scale Models (≥70B)2026.04 | 43.75 | 46.81 | 69.98 | 34.32 | 29.17 | 8.31 | 62.63 | 59.92 | 66.34 | 41.03 | 46.23 | |
| Qwen3-VL-8BModel Scale=7 ∼ 8B Models2026.04 | 41.1 | 23.4 | 68.63 | 34.59 | 31.94 | 38.73 | 52 | 60.16 | 63.37 | 22.65 | 43.66 | |
| Visual-RFT-7BBase MLLM=Qwen2-VL2026.04 | 33.95 | 31.29 | 50 | 15.36 | 12.21 | 14.98 | 24.25 | 62.72 | 41.89 | 10.4 | 29.71 | |
| Qwen2-VL-7BModel Scale=7 ∼ 8B Models2026.04 | 27.84 | 38.3 | 19.36 | 20.73 | 11.81 | 25.95 | 23.23 | 58.52 | 48.51 | 11.97 | 28.62 | |
| InternVL3-8BModel Scale=7 ∼ 8B Models2026.04 | 23.67 | 14.89 | 47.99 | 14.84 | 6.94 | 12.13 | 7.07 | 34.87 | 16.83 | 2.56 | 18.18 | |
| mPLUG-Owl3-8BModel Scale=7 ∼ 8B Models2026.04 | 18.56 | 6.38 | 34.93 | 8.55 | 7.64 | 2.41 | 7.07 | 22.85 | 9.09 | 5.98 | 12.35 | |
| InternVL2-76BModel Scale=Large-Scale Models (≥70B)2026.04 | 15.91 | 10.64 | 36.4 | 30.73 | 20.83 | 5.74 | 46.46 | 41.28 | 32.67 | 26.5 | 26.72 | |
| MiniCPM2.6-8BModel Scale=7 ∼ 8B Models2026.04 | 14.58 | 2.13 | 14.34 | 9.82 | 6.25 | 1.75 | 11.11 | 10.02 | 2.97 | 2.56 | 7.55 | |
| LLaVA-OV-72BModel Scale=Large-Scale Models (≥70B)2026.04 | 13.26 | 5.34 | 26.84 | 12.91 | 7.64 | 2.14 | 17.83 | 21.6 | 11.88 | 8.55 | 13.65 | |
| InternVL3-78BModel Scale=Large-Scale Models (≥70B)2026.04 | 10.04 | 9.57 | 24.12 | 27.08 | 14.58 | 10.44 | 50.51 | 38.08 | 45.54 | 17.09 | 24.71 | |
| Qwen2.5-VL-7BModel Scale=7 ∼ 8B Models2026.04 | 9.57 | 21.59 | 22.42 | 10.58 | 5.88 | 1.25 | 10 | 33.47 | 2.97 | 2.6 | 12.03 | |
| InternVL2-8BModel Scale=7 ∼ 8B Models2026.04 | 6.92 | 7.45 | 25.49 | 20.73 | 9.72 | 3.49 | 28.28 | 30.26 | 17.82 | 9.4 | 15.96 | |
| LLaVA-OV-7BModel Scale=7 ∼ 8B Models2026.04 | 6.06 | 3.19 | 3.43 | 0.18 | 1.04 | 1.08 | 9.09 | 15.43 | 6.93 | 0.85 | 4.73 | |
| Mantis-8BModel Scale=7 ∼ 8B Models2026.04 | 1.52 | 0 | 3.31 | 12.18 | 2.08 | 1 | 1.01 | 10.02 | 0 | 0.85 | 3.2 |