Visual Logical Reasoning on VisuLogic
28.5AccuracyERNIE 5.0-Base
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ERNIE 5.0-BaseModel type=pre-trained2026.02 | 28.5 | — | — | |
| SketchThinker-R1-3BBackbone=Qwen2.5-VL-3B-Instruct, Paradigm=Reinforcement-Learning-based2026.01 | 25.8 | 36.9 | 0.699 | |
| Vanilla-R1Backbone=Qwen2.5-VL-3B-Instruct, Paradigm=Direct Inference2026.01 | 25.6 | 139.5 | 0.184 | |
| ThinkPruneBackbone=Qwen2.5-VL-3B-Instruct, Paradigm=Reinforcement-Learning-based2026.01 | 25.4 | 73.3 | 0.347 | |
| Constrained CoTBackbone=Qwen2.5-VL-3B-Instruct, Paradigm=Prompt-based2026.01 | 25.2 | 69.1 | 0.365 | |
| VeriThinkerBackbone=Qwen2.5-VL-3B-Instruct, Paradigm=Supervised-fine-tuning-based2026.01 | 25.2 | 93.7 | 0.269 | |
| L1Backbone=Qwen2.5-VL-3B-Instruct, Paradigm=Reinforcement-Learning-based2026.01 | 25.2 | 97.6 | 0.258 | |
| Chain-of-DraftBackbone=Qwen2.5-VL-3B-Instruct, Paradigm=Prompt-based2026.01 | 25.1 | 74.2 | 0.338 | |
| C3oTBackbone=Qwen2.5-VL-3B-Instruct, Paradigm=Supervised-fine-tuning-based2026.01 | 25.1 | 104.3 | 0.241 | |
| EVE-iter3Backbone=MiMo-VL-7B-SFT-2508, Training Iteration=32026.04 | 23.7 | — | — | |
| EVE-iter2Backbone=MiMo-VL-7B-SFT-2508, Training Iteration=22026.04 | 23.4 | — | — | |
| EVE-iter1Backbone=MiMo-VL-7B-SFT-2508, Training Iteration=12026.04 | 22.1 | — | — | |
| MiMo-VL-7B-SFT-2508Backbone=MiMo-VL-7B-SFT-2508, Training Iteration=02026.04 | 20.3 | — | — |