Real-world QA on RealworldQA
81.8AccuracyM2Note
Evaluation Results
| Method | Links | |
|---|---|---|
| M2NoteBackbone=GPT-5.4, Evolving Mode=Self-Evolving, Mem=28, Len=4622026.07 | 81.8 | |
| VanillaBackbone=GPT-5.4, Evolving Mode=Self-Evolving2026.07 | 80.8 | |
| M2NoteBackbone=Qwen3-VL-Plus, Evolving Mode=Self-Evolving, Mem=28, Len=4622026.07 | 79.7 | |
| CoT+M2NoteBackbone=Qwen3-VL-8B-Instruct, Evolving Mode=Self-Evolving, Mem=32, Len=3152026.07 | 78.8 | |
| VanillaBackbone=Qwen3-VL-Plus, Evolving Mode=Self-Evolving2026.07 | 78.2 | |
| CoTBackbone=Qwen3-VL-8B-Instruct, Evolving Mode=Self-Evolving2026.07 | 78 | |
| M2NoteBackbone (Tuning)=Qwen3-VL-8B-Instruct, Backbone (Tuner)=Qwen3-VL-Plus, Evolving Mode=Cross-Model Evolving, Mem=43, Len=4452026.07 | 77.1 | |
| Qwen2.5-VL-72B2026.04 | 75.7 | |
| GPT-4o2026.04 | 75.4 | |
| M2NoteBackbone=Qwen3-VL-8B-Instruct, Evolving Mode=Self-Evolving, Mem=50, Len=2612026.07 | 75.4 | |
| Qwen2.5-VL-7B AIFmethod=Adaptive Information Flow2026.04 | 74.5 | |
| VanillaBackbone=Qwen3-VL-8B-Instruct, Evolving Mode=Self-Evolving2026.07 | 73.2 | |
| VanillaBackbone (Tuning)=Qwen3-VL-8B-Instruct, Backbone (Tuner)=Qwen3-VL-Plus, Evolving Mode=Cross-Model Evolving2026.07 | 73.2 | |
| SwimBird2026.02 | 73.1 | |
| S^3-FTBackbone=Qwen3-VL-8B, Mode=Single-view2026.05 | 72.2 | |
| DPEBackbone=Qwen3-VL-8B-Instruct, Evolving Mode=Self-Evolving2026.07 | 72.1 | |
| LLaVA-OneVision-72BModel Scale=72B2024.08 | 71.9 | |
| SAIL-VLModel Size=8B2025.01 | 71.9 | |
| Qwen3-VL-8B-Instructreproduced by authors=true2026.02 | 71.8 | |
| LFM2-VL-3B# Total Params=3.0B, Evaluation Toolkit=VLMEvalKit2025.11 | 71.37 | |
| Qwen3-VL-8BMode=base2026.05 | 71 | |
| InternVL2.5-MPOModel Size=8B2025.01 | 70.2 | |
| DeepSeekVL-2Model Size=8B2025.01 | 70.2 | |
| LLaVA-OneVision2026.02 | 69.9 | |
| Qwen2-VLModel Size=8B2025.01 | 69.7 | |
| SkiLa2026.02 | 69.3 | |
| S^3-FTBackbone=Qwen3-VL-8B, Mode=Cross-viewpoint2026.05 | 69 | |
| VisPlayEvolving Mode=Self-Evolving2026.07 | 69 | |
| S^3-FTBackbone=Qwen2.5-VL-7B, Mode=Single-view2026.05 | 68.8 | |
| Qwen2.5-VL-7B2026.04 | 68.5 | |
| Vision-ZeroEvolving Mode=Self-Evolving2026.07 | 68.5 | |
| Qwen2.5-VL-7BMode=base2026.05 | 68.2 | |
| Qwen2.5-VL-7B-Instruct2026.02 | 67.4 | |
| InfiniteVL-4BComplexity=O(1)2025.12 | 67.3 | |
| S^3-FTBackbone=Qwen2.5-VL-7B, Mode=Cross-viewpoint2026.05 | 66.7 | |
| LLaVA-OneVision-7BModel Scale=7B2024.08 | 66.3 | |
| Qwen3-VL-2B# Total Params=2.1B, Evaluation Toolkit=VLMEvalKit2025.11 | 65.75 | |
| InternVL3.5-8BRatio=100%, Backbone=InternVL3.5-8B2026.05 | 65.62 | |
| Qwen2.5VL-3BComplexity=O(n)2025.12 | 65.4 | |
| Qwen2.5-VL-3B# Total Params=3.75B, Evaluation Toolkit=VLMEvalKit2025.11 | 65.23 | |
| DeepSeekVL-2Model Size=2B2025.01 | 64.8 | |
| InternVL2.5-4BComplexity=O(n)2025.12 | 64.3 | |
| F3ARatio=60%, Backbone=InternVL3.5-8B2026.05 | 63.98 | |
| SAIL-VLModel Size=2B2025.01 | 63.8 | |
| VisionZipRatio=60%, Backbone=InternVL3.5-8B2026.05 | 63.39 | |
| CDPrunerRatio=60%, Backbone=InternVL3.5-8B2026.05 | 63.34 | |
| DivPruneRatio=60%, Backbone=InternVL3.5-8B2026.05 | 63.26 | |
| CDPrunerRatio=40%, Backbone=InternVL3.5-8B2026.05 | 62.88 | |
| CASA⊕ Qwen2.5-VLRes.=≤ 896²2025.12 | 62.5 | |
| FastVRatio=60%, Backbone=InternVL3.5-8B2026.05 | 62.27 | |
| F3ARatio=40%, Backbone=InternVL3.5-8B2026.05 | 62.26 | |
| Qwen2.5-VL 3B (reproduced)Res.=≤ 896²2025.12 | 62.2 | |
| VisionZipRatio=40%, Backbone=InternVL3.5-8B2026.05 | 61.98 | |
| DivPruneRatio=40%, Backbone=InternVL3.5-8B2026.05 | 61.68 | |
| GPT-4VVersion=V-Preview2024.08 | 61.4 | |
| FastVRatio=40%, Backbone=InternVL3.5-8B2026.05 | 61.03 | |
| Qwen2-VLModel Size=2B2025.01 | 60.9 | |
| InternVL3.5-2B# Total Params=2.3B, Evaluation Toolkit=VLMEvalKit2025.11 | 60.78 | |
| InternVL2.5-MPOModel Size=2B2025.01 | 60.7 | |
| LLaVA-1.5-7B AIFmethod=Adaptive Information Flow2026.04 | 60.5 | |
| Claude-3.5 Sonnet2026.04 | 60.1 | |
| F3ARatio=20%, Backbone=InternVL3.5-8B2026.05 | 59.85 | |
| DivPruneRatio=20%, Backbone=InternVL3.5-8B2026.05 | 59.39 | |
| SmolVLM2-2BComplexity=O(n)2025.12 | 58.4 | |
| PaliGemma2-3BComplexity=O(n)2025.12 | 58.3 | |
| VisionZipRatio=20%, Backbone=InternVL3.5-8B2026.05 | 58.07 | |
| Phi-3.5-Vision-4BComplexity=O(n)2025.12 | 57.9 | |
| SmolVLM2-2.2B# Total Params=2.2B, Evaluation Toolkit=VLMEvalKit2025.11 | 57.5 | |
| LLaVA-OneVision-0.5BModel Scale=0.5B2024.08 | 55.6 | |
| LLaVA-1.5-7B2026.04 | 55.6 | |
| CDPrunerRatio=20%, Backbone=InternVL3.5-8B2026.05 | 55.42 | |
| TinyLLaVA-3BComplexity=O(n)2025.12 | 55 | |
| FastVRatio=20%, Backbone=InternVL3.5-8B2026.05 | 52.63 | |
| MaTVLM-3BComplexity=O(n)2025.12 | 52.3 | |
| Cobra-3BComplexity=O(1)2025.12 | 51 |