Visual Reasoning on BLINK-J
88AccuracyAdaReasoner 7B
Evaluation Results
| Method | Links | |
|---|---|---|
| AdaReasoner 7BModel Category=Tool-Planning Models, Backbone=7B2026.01 | 88 | |
| GPT 5 + ToolsModel Category=Tool-Planning Models, Tool Usage=true2026.01 | 76 | |
| ProLaViT (Full)Reasoning Paradigm=Progressive Latent Visual Thought, Reasoning Space=Latent, Training Strategy=Distance-Weighted Diversity Loss2026.07 | 76 | |
| GPT 5Model Category=Closed-Source Models2026.01 | 73.33 | |
| Qwen 2.5 VL 72BModel Category=Open-Source Models, Backbone=72B2026.01 | 71.33 | |
| ProLaViT (Base)Reasoning Paradigm=Progressive Latent Visual Thought, Reasoning Space=Latent2026.07 | 71.33 | |
| CoT SFTReasoning Paradigm=Text-Space Reasoning, Reasoning Space=Text2026.07 | 68.66 | |
| ProLaViT + L_div^marginReasoning Paradigm=Progressive Latent Visual Thought, Reasoning Space=Latent2026.07 | 67.33 | |
| CoVTReasoning Paradigm=Latent-Space Reasoning, Reasoning Space=Latent2026.07 | 66 | |
| One-step Latent Pred.Reasoning Paradigm=Latent-Space Reasoning, Reasoning Space=Latent2026.07 | 66 | |
| Gemini 2.5 flashModel Category=Closed-Source Models2026.01 | 65.33 | |
| Claude 4 sonnetModel Category=Closed-Source Models2026.01 | 65.33 | |
| Qwen 2.5 VL 72B + ToolsModel Category=Tool-Planning Models, Tool Usage=true, Backbone=72B2026.01 | 65.33 | |
| Qwen 2.5 VL 32BModel Category=Open-Source Models, Backbone=32B2026.01 | 64.67 | |
| InternVL3 78BModel Category=Open-Source Models, Backbone=78B2026.01 | 60 | |
| PixelReasonerModel Category=Tool-Planning Models, Unified Evaluation Framework=true2026.01 | 59.33 | |
| Qwen2.5-VL-InstructReasoning Paradigm=Base Model, Reasoning Space=N/A2026.07 | 59.33 | |
| LVRReasoning Paradigm=Latent-Space Reasoning, Reasoning Space=Latent2026.07 | 57.33 | |
| Qwen 2.5 VL 7B + ToolsModel Category=Tool-Planning Models, Tool Usage=true, Backbone=7B2026.01 | 56 | |
| Qwen 2.5 VL 7BModel Category=Open-Source Models, Backbone=7B2026.01 | 52.67 | |
| DeepEyesModel Category=Tool-Planning Models, Unified Evaluation Framework=true2026.01 | 50.67 | |
| SFTReasoning Paradigm=Text-Space Reasoning, Reasoning Space=Text2026.07 | 49.33 | |
| Qwen 2.5 VL 3BModel Category=Open-Source Models, Backbone=3B2026.01 | 48.67 |