Visual Grounded Reasoning on TreeBench
54.8Overall Scoreo3-0416
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| o3-0416Model Category=Private Model2025.11 | 54.8 | — | 69 | 69.2 | 65.2 | 68.8 | 79.4 | 22.4 | 38.6 | 61 | 86.2 | 50 | — | — | |
| GPT-o3-04162026.03 | 54.8 | — | 69 | 69.2 | 65.2 | 68.8 | 79.4 | 22.4 | 38.6 | 61 | 86.2 | 50 | — | — | |
| o3-0416Model Type=Private Models2025.07 | 54.8 | — | 69 | 69.2 | 65.2 | 68.8 | 79.4 | 22.4 | 38.6 | 61 | 86.2 | 50 | — | — | |
| o3-04162026.02 | 54.8 | — | 69 | 69.2 | 65.2 | 68.8 | 79.4 | 22.4 | 38.6 | 61 | 86.2 | 50 | — | — | |
| o3Parameters=—2026.02 | 54.8 | — | 69 | 69.2 | 65.2 | 68.8 | 79.4 | 22.4 | 38.6 | 61 | 86.2 | 50 | — | — | |
| Gemini-2.5-Pro-0605Model Category=Private Model2025.11 | 54.1 | — | 51.7 | 61.5 | 56.5 | 75 | 83.8 | 20 | 36.8 | 65.9 | 86.2 | 54.6 | — | — | |
| Gemini-2.5-Pro-06052026.03 | 54.1 | — | 51.7 | 61.5 | 56.5 | 75 | 83.8 | 20 | 36.8 | 65.9 | 86.2 | 54.6 | — | — | |
| Gemini-2.5-Pro-0605Model Type=Private Models2025.07 | 54.1 | — | 51.7 | 61.5 | 56.5 | 75 | 83.8 | 20 | 36.8 | 65.9 | 86.2 | 54.6 | — | — | |
| Gemini-2.5-Pro-06052026.02 | 54.1 | — | 51.7 | 61.5 | 56.5 | 75 | 83.8 | 20 | 36.8 | 65.9 | 86.2 | 54.6 | — | — | |
| Gemini-2.5-ProParameters=—2026.02 | 54.1 | — | 51.7 | 61.5 | 56.5 | 75 | 83.8 | 20 | 36.8 | 65.9 | 86.2 | 54.6 | — | — | |
| TTSPBackbone=Qwen3-VL-8B-Instruct2026.04 | 51.8 | — | — | — | — | — | — | — | — | — | — | — | 73.2 | 39.4 | |
| Nemotron 3 Nano OmniOpen-Source=true, Size=30B-A3B, Reasoning=on2026.04 | 51.6 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GRiPModel Category=Open-source Visual Grounded Reasoning Model2025.11 | 51.3 | 45 | 69.1 | 54.4 | 83.4 | 69.3 | 64.1 | 23.2 | 38.6 | 58.7 | 69.4 | 45.8 | — | — | |
| InternVL3-78BECRD=true2026.02 | 50.9 | — | 65.5 | 69.2 | 65.2 | 75 | 60.3 | 21.2 | 35.1 | 63.4 | 86.2 | 47.7 | — | — | |
| TTSPBackbone=Qwen3-VL-4B-Instruct2026.04 | 50.9 | — | — | — | — | — | — | — | — | — | — | — | 70.5 | 39.4 | |
| TreeVGR-7BModel Category=Open-source Visual Grounded Reasoning Model, Parameters=7B2025.11 | 50.4 | 44 | 65.5 | 53.8 | 82.6 | 68.8 | 63.3 | 22.4 | 36.8 | 61 | 69 | 45.5 | — | — | |
| TreeVGR-7BModel Type=Open-source Visual Grounded Reasoning Models, Parameters=7B2025.07 | 50.4 | 44 | 65.5 | 53.8 | 82.6 | 68.8 | 63.3 | 22.4 | 36.8 | 61 | 69 | 45.5 | — | — | |
| TreeVGR-7B2026.02 | 50.4 | — | 65.5 | 53.8 | 82.6 | 68.8 | 63.3 | 22.4 | 36.8 | 61 | 69 | 45.5 | — | — | |
| RTWIBackbone=Qwen3-VL-8B-Instruct2026.04 | 50.1 | — | — | — | — | — | — | — | — | — | — | — | 71.1 | 37.9 | |
| Qwen2.5-VL-72BECRD=true2026.02 | 49.9 | — | 65.5 | 69.2 | 65.2 | 62.5 | 69.1 | 20 | 36.8 | 58.5 | 75.9 | 40.9 | — | — | |
| DeepconfBackbone=Qwen3-VL-8B-Instruct2026.04 | 49.9 | — | — | — | — | — | — | — | — | — | — | — | 70.4 | 37.9 | |
| TreeVGRPixel Constraint=640x28x28, Post-training Category=Non-efficiency-oriented Post Training Methods, Inference Pipeline=inference with original pipeline2025.09 | 49.1 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| RTWIBackbone=Qwen3-VL-4B-Instruct2026.04 | 49.1 | — | — | — | — | — | — | — | — | — | — | — | 66.4 | 39 | |
| SCBackbone=Qwen3-VL-8B-Instruct2026.04 | 49.1 | — | — | — | — | — | — | — | — | — | — | — | 69.8 | 37.1 | |
| TRTBackbone=Qwen3-VL-8B-Instruct2026.04 | 48.9 | — | — | — | — | — | — | — | — | — | — | — | 69.1 | 37.1 | |
| Qwen2.5-VL-32BECRD=true2026.02 | 48.6 | — | 62.1 | 53.8 | 73.9 | 62.5 | 60.3 | 23.5 | 35.1 | 61 | 65.5 | 45.5 | — | — | |
| Self-Ref.Backbone=Qwen3-VL-8B-Instruct2026.04 | 48.6 | — | — | — | — | — | — | — | — | — | — | — | 69.1 | 36.7 | |
| InternVL3-38BECRD=true2026.02 | 48.4 | — | 51.7 | 69.2 | 65.2 | 75 | 58.8 | 22.4 | 36.8 | 56.1 | 69 | 50 | — | — | |
| TreeVGRPixel Constraint=1280x28x28, Post-training Category=Non-efficiency-oriented Post Training Methods, Inference Pipeline=inference with original pipeline2025.09 | 48.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepconfBackbone=Qwen3-VL-4B-Instruct2026.04 | 48.2 | — | — | — | — | — | — | — | — | — | — | — | 65.8 | 37.9 | |
| CISCBackbone=Qwen3-VL-8B-Instruct2026.04 | 48.2 | — | — | — | — | — | — | — | — | — | — | — | 67.8 | 36.7 | |
| Self-Cer.Backbone=Qwen3-VL-8B-Instruct2026.04 | 48.2 | — | — | — | — | — | — | — | — | — | — | — | 67.8 | 36.7 | |
| Qwen2.5-VL-7BECRD=true2026.02 | 47.9 | — | 72.4 | 53.8 | 73.9 | 62.5 | 54.4 | 20 | 35.1 | 56.1 | 75.9 | 45.5 | — | — | |
| SCBackbone=Qwen3-VL-4B-Instruct2026.04 | 47.7 | — | — | — | — | — | — | — | — | — | — | — | 64.4 | 37.9 | |
| Self-Ref.Backbone=Qwen3-VL-4B-Instruct2026.04 | 47.7 | — | — | — | — | — | — | — | — | — | — | — | 64.4 | 37.9 | |
| CISCBackbone=Qwen3-VL-4B-Instruct2026.04 | 47.2 | — | — | — | — | — | — | — | — | — | — | — | 63.8 | 37.5 | |
| TRTBackbone=Qwen3-VL-4B-Instruct2026.04 | 47.2 | — | — | — | — | — | — | — | — | — | — | — | 63.8 | 37.5 | |
| Self-Cer.Backbone=Qwen3-VL-4B-Instruct2026.04 | 47.2 | — | — | — | — | — | — | — | — | — | — | — | 63.7 | 37.5 | |
| GPT-4o-1120Model Category=Private Model2025.11 | 46.9 | — | 51.7 | 61.5 | 65.2 | 43.8 | 69.1 | 18.8 | 38.6 | 48.8 | 72.4 | 43.2 | — | — | |
| GPT-4o-11202026.03 | 46.9 | — | 51.7 | 61.5 | 65.2 | 43.8 | 69.1 | 18.8 | 38.6 | 48.8 | 72.4 | 43.2 | — | — | |
| GPT-4o-1120Model Type=Private Models2025.07 | 46.9 | — | 51.7 | 61.5 | 65.2 | 43.8 | 69.1 | 18.8 | 38.6 | 48.8 | 72.4 | 43.2 | — | — | |
| GPT-4o-11202026.02 | 46.9 | — | 51.7 | 61.5 | 65.2 | 43.8 | 69.1 | 18.8 | 38.6 | 48.8 | 72.4 | 43.2 | — | — | |
| LLaVA-OneVision-72BECRD=true2026.02 | 46.9 | — | 62.1 | 61.5 | 69.6 | 68.8 | 51.5 | 18.8 | 29.8 | 61 | 72.4 | 52.3 | — | — | |
| GPT-4oParameters=—2026.02 | 46.9 | — | 51.7 | 61.5 | 65.2 | 43.8 | 69.1 | 18.8 | 38.6 | 48.8 | 72.4 | 43.2 | — | — | |
| GPT-4o2026.04 | 46.9 | — | — | — | — | — | — | — | — | — | — | — | 61.7 | 38.3 | |
| GPT-4oVisual Bandwidth Constraint (B)=No Constrain, Reported From=Other publications2026.05 | 46.9 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| GPT-4o2026.06 | 46.9 | — | 51.7 | 61.5 | 65.2 | 43.8 | 69.1 | 18.8 | 38.6 | 48.8 | 72.4 | 43.2 | — | — | |
| InternVL3-78BModel Category=Open-source General Model, Parameters=78B2025.11 | 46.4 | — | 62.1 | 61.5 | 52.2 | 68.8 | 52.9 | 16.5 | 33.3 | 61 | 86.2 | 45.5 | — | — | |
| InternVL3-78BParameter Scale=78B2026.03 | 46.4 | — | 62.1 | 61.5 | 52.2 | 68.8 | 52.9 | 16.5 | 33.3 | 61 | 86.2 | 45.5 | — | — | |
| InternVL3-78BModel Type=Open-source General Models, Parameters=78B2025.07 | 46.4 | — | 62.1 | 61.5 | 52.2 | 68.8 | 52.9 | 16.5 | 33.3 | 61 | 86.2 | 45.5 | — | — | |
| InternVL3-78BECRD=false2026.02 | 46.4 | — | 62.1 | 61.5 | 52.2 | 68.8 | 52.9 | 16.5 | 33.3 | 61 | 86.2 | 45.5 | — | — | |
| BaseBackbone=Qwen3-VL-8B-Instruct2026.04 | 46.4 | — | — | — | — | — | — | — | — | — | — | — | 67.1 | 34.4 | |
| Gemini-2.5-Flash-0520Model Category=Private Model2025.11 | 45.9 | — | 48.3 | 53.9 | 69.6 | 68.8 | 75 | 15.3 | 19.3 | 56.1 | 72.4 | 43.2 | — | — | |
| Gemini-2.5-Flash-05202026.03 | 45.9 | — | 48.3 | 53.9 | 69.6 | 68.8 | 75 | 15.3 | 19.3 | 56.1 | 72.4 | 43.2 | — | — | |
| Gemini-2.5-Flash-0520Model Type=Private Models2025.07 | 45.9 | — | 48.3 | 53.9 | 69.6 | 68.8 | 75 | 15.3 | 19.3 | 56.1 | 72.4 | 43.2 | — | — | |
| Gemini-2.5-Flash-05202026.02 | 45.9 | — | 48.3 | 53.9 | 69.6 | 68.8 | 75 | 15.3 | 19.3 | 56.1 | 72.4 | 43.2 | — | — | |
| Gemini-2.5-FlashParameters=—2026.02 | 45.9 | — | 48.3 | 53.9 | 69.6 | 68.8 | 75 | 15.3 | 19.3 | 56.1 | 72.4 | 43.2 | — | — | |
| InternVL3-8BECRD=true2026.02 | 45.2 | — | 51.7 | 69.2 | 69.6 | 62.5 | 50 | 24.7 | 29.8 | 43.9 | 72.4 | 50 | — | — | |
| BaseBackbone=Qwen3-VL-4B-Instruct2026.04 | 45.2 | — | — | — | — | — | — | — | — | — | — | — | 60.4 | 36.3 | |
| HART-7BParameters=7B2026.02 | 43.7 | — | 62.1 | 61.5 | 52.2 | 56.3 | 55.9 | 14.1 | 35.1 | 48.8 | 62.1 | 50 | — | — | |
| Nemotron 3 Nano OmniOpen-Source=true, Size=30B-A3B, Reasoning=off2026.04 | 43.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LLaVA-OneVision-7BECRD=true2026.02 | 43.5 | — | 55.2 | 61.5 | 60.9 | 56.3 | 47.1 | 23.5 | 28 | 53.7 | 72.4 | 40.9 | — | — | |
| LazyMCoTBackbone=Qwen3-VL-8B-Instruct2026.06 | 43.5 | — | 55.2 | 61.5 | 65.2 | 75 | 45.6 | 16.5 | 29.8 | 48.8 | 72.4 | 50 | — | — | |
| ERGOPixel Constraint=640x28x28, Post-training Category=Efficiency-oriented Post Training Methods2025.09 | 43.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| BaseBackbone=Qwen3-VL-8B-Instruct2026.06 | 43 | — | 55.2 | 61.5 | 69.6 | 75 | 45.6 | 18.8 | 28.1 | 41.5 | 69 | 50 | — | — | |
| Qwen2.5-VL-32BModel Category=Open-source General Model, Parameters=32B2025.11 | 42.5 | — | 51.7 | 53.8 | 69.6 | 62.5 | 54.4 | 16.5 | 33.3 | 46.3 | 62.1 | 38.6 | — | — | |
| Qwen2.5-VL-32BParameter Scale=32B2026.03 | 42.5 | — | 51.7 | 53.8 | 69.6 | 62.5 | 54.4 | 16.5 | 33.3 | 46.3 | 62.1 | 38.6 | — | — | |
| DeepScanRL-based=false, Backbone=Qwen2.5-VL-7B2026.03 | 42.5 | 37.3 | 62.1 | 69.2 | 60.9 | 68.8 | 44.1 | 21.2 | 36.8 | 43.9 | 48.3 | 43.2 | — | — | |
| Qwen2.5-VL-32BECRD=false2026.02 | 42.5 | — | 51.7 | 53.8 | 69.6 | 62.5 | 54.4 | 16.5 | 33.3 | 46.3 | 62.1 | 38.6 | — | — | |
| Qwen2.5-VL-32BParameters=32B2026.02 | 42.5 | — | 51.7 | 53.8 | 69.6 | 62.5 | 54.4 | 16.5 | 33.3 | 46.3 | 62.1 | 38.6 | — | — | |
| Nemotron Nano V2 VLOpen-Source=true, Size=12B, Reasoning=on2026.04 | 42.5 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| PixelReasonerPixel Constraint=640x28x28, Post-training Category=Non-efficiency-oriented Post Training Methods2025.09 | 42.3 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-VL-72BModel Category=Open-source General Model, Parameters=72B2025.11 | 42.2 | — | 65.5 | 69.2 | 56.5 | 56.3 | 48.5 | 11.8 | 33.3 | 51.2 | 72.4 | 38.6 | — | — | |
| Qwen2.5-VL-72BParameter Scale=72B2026.03 | 42.2 | — | 65.5 | 69.2 | 56.5 | 56.3 | 48.5 | 11.8 | 33.3 | 51.2 | 72.4 | 38.6 | — | — | |
| Qwen2.5-VL-72BModel Type=Open-source General Models, Parameters=72B2025.07 | 42.2 | — | 65.5 | 69.2 | 56.5 | 56.3 | 48.5 | 11.8 | 33.3 | 51.2 | 72.4 | 38.6 | — | — | |
| Qwen2.5-VL-72BECRD=false2026.02 | 42.2 | — | 65.5 | 69.2 | 56.5 | 56.3 | 48.5 | 11.8 | 33.3 | 51.2 | 72.4 | 38.6 | — | — | |
| Qwen2.5-VL-72BParameters=72B2026.02 | 42.2 | — | 65.5 | 69.2 | 56.5 | 56.3 | 48.5 | 11.8 | 33.3 | 51.2 | 72.4 | 38.6 | — | — | |
| InternVL3-38BModel Category=Open-source General Model, Parameters=38B2025.11 | 42 | — | 51.7 | 61.5 | 52.2 | 68.8 | 51.5 | 12.9 | 33.3 | 56.1 | 65.5 | 38.6 | — | — | |
| InternVL3-38BParameter Scale=38B2026.03 | 42 | — | 51.7 | 61.5 | 52.2 | 68.8 | 51.5 | 12.9 | 33.3 | 56.1 | 65.5 | 38.6 | — | — | |
| InternVL3-38BECRD=false2026.02 | 42 | — | 51.7 | 61.5 | 52.2 | 68.8 | 51.5 | 12.9 | 33.3 | 56.1 | 65.5 | 38.6 | — | — | |
| InternVL3-38BParameters=38B2026.02 | 42 | — | 51.7 | 61.5 | 52.2 | 68.8 | 51.5 | 12.9 | 33.3 | 56.1 | 65.5 | 38.6 | — | — | |
| MGPOPixel Constraint=640x28x28, Post-training Category=Efficiency-oriented Post Training Methods, Inference Pipeline=reproduction with their code using our data2025.09 | 42 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| ERGOPixel Constraint=1280x28x28, Post-training Category=Efficiency-oriented Post Training Methods2025.09 | 41.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LazyMCoTBackbone=Qwen2.5-VL-7B-Instruct2026.06 | 41.7 | — | 58.6 | 61.5 | 56.5 | 68.8 | 39.7 | 21.2 | 31.6 | 46.3 | 48.3 | 54.5 | — | — | |
| HiDeBackbone=Qwen3-VL-8B-Instruct2026.06 | 41.7 | — | 48.3 | 46.2 | 65.2 | 75 | 45.6 | 15.3 | 24.6 | 53.7 | 65.5 | 52.3 | — | — | |
| TreeVGRRL-based=true, Backbone=Qwen2.5-VL-7B2026.03 | 41 | 31.8 | 55.2 | 61.5 | 65.2 | 50 | 61.7 | 15.3 | 24.6 | 46.3 | 44.8 | 40.9 | — | — | |
| VisionThinkPixel Constraint=1280x28x28, Post-training Category=Efficiency-oriented Post Training Methods, Inference Pipeline=inference with original pipeline2025.09 | 41 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| CVSearchBackbone=Qwen2.5-VL-7B2026.05 | 40.7 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| LLaVA-OneVision-72BModel Category=Open-source General Model, Parameters=72B2025.11 | 40.5 | — | 62.1 | 53.8 | 65.2 | 62.3 | 36.8 | 12.9 | 28.1 | 53.7 | 65.5 | 47.7 | — | — | |
| LLaVA-OneVision-72BParameter Scale=72B2026.03 | 40.5 | — | 62.1 | 53.8 | 65.2 | 62.3 | 36.8 | 12.9 | 28.1 | 53.7 | 65.5 | 47.7 | — | — | |
| LLaVA-OneVision-72BModel Type=Open-source General Models, Parameters=72B2025.07 | 40.5 | — | 62.1 | 53.8 | 65.2 | 62.3 | 36.8 | 12.9 | 28.1 | 53.7 | 65.5 | 47.7 | — | — | |
| LLaVA-OneVision-72BECRD=false2026.02 | 40.5 | — | 62.1 | 53.8 | 65.2 | 62.5 | 36.8 | 12.9 | 28.1 | 53.7 | 65.5 | 47.7 | — | — | |
| LLaVA-OneVision-72BParameters=72B2026.02 | 40.5 | — | 62.1 | 53.8 | 65.2 | 62.3 | 36.8 | 12.9 | 28.1 | 53.7 | 65.5 | 47.7 | — | — | |
| PixelReasonerPixel Constraint=1280x28x28, Post-training Category=Non-efficiency-oriented Post Training Methods2025.09 | 40.4 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| MGPO + SFTParameters=7B2026.02 | 40.3 | — | 58.6 | 61.5 | 52.2 | 56.3 | 48.5 | 14.1 | 31.6 | 51.2 | 51.7 | 40.9 | — | — | |
| Qwen2.5-VL + BA-SFT + VanillaRLVisual Bandwidth Constraint (B)=No Constrain, Tool-calling=True2026.05 | 40.25 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-VL + BA-SFTVisual Bandwidth Constraint (B)=No Constrain, Tool-calling=True2026.05 | 40.24 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| DeepEyesPixel Constraint=1280x28x28, Post-training Category=Non-efficiency-oriented Post Training Methods2025.09 | 40.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-VL-7B-Inst.Pixel Constraint=640x28x282025.09 | 40.2 | — | — | — | — | — | — | — | — | — | — | — | — | — | |
| Qwen2.5-VL-7B-Inst.Pixel Constraint=16384x28x282025.09 | 40 | — | — | — | — | — | — | — | — | — | — | — | — | — |