Compositional Reasoning on BLINK
68AccuracyGPT-4o
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-4oParadigm=Zero-Shot VLMs, Data Size=-2026.04 | 68 | |
| SCF-VRParadigm=Latent Reasoning, Data Size=30K, Backbone=Qwen2.5-VL-7B2026.04 | 57.96 | |
| DMLRParadigm=Latent Reasoning, Data Size=-2026.04 | 56.92 | |
| VL-RethinkerParadigm=Tool-use & RL Enhanced Reasoning, Data Size=39K2026.04 | 55.55 | |
| InternVL3.5-8BParadigm=Zero-Shot VLMs, Data Size=70K2026.04 | 54.81 | |
| Qwen2.5-VL-7BParadigm=Zero-Shot VLMs, Data Size=-2026.04 | 53.6 | |
| LVRParadigm=Latent Reasoning, Data Size=470K2026.04 | 53.6 | |
| Vision-R1Paradigm=Tool-use & RL Enhanced Reasoning, Data Size=200K2026.04 | 52.71 | |
| PAPOParadigm=Tool-use & RL Enhanced Reasoning, Data Size=39K2026.04 | 52.66 | |
| DeepEyesParadigm=Tool-use & RL Enhanced Reasoning, Data Size=47K2026.04 | 51.08 | |
| MonetParadigm=Latent Reasoning, Data Size=125K2026.04 | 50.71 | |
| LLaVA-OneVisionParadigm=Zero-Shot VLMs, Data Size=9M2026.04 | 49.34 |