Perception-intensive Reasoning on MMStar
64AccuracyVL-Rethinker
Evaluation results
| Method | Links | |
|---|---|---|
| VL-Rethinker#Param=7B2026.04 | 64 | |
| VLAA-Thinker#Param=7B2026.04 | 64 | |
| Perceval#Param=7B2026.04 | 63.8 | |
| MM-Eureka#Param=7B2026.04 | 63.4 | |
| Pixel-Reasoner#Param=7B, Tool Capability=true2026.04 | 63.1 | |
| OpenVLThinker#Param=7B2026.04 | 62.8 | |
| DeepEyes#Param=7B, Tool Capability=true2026.04 | 62.7 | |
| Qwen2.5-VL#Param=7B2026.04 | 62.3 | |
| Qwen2.5-VL + GRPO#Param=7B2026.04 | 62 | |
| VLM-R1#Param=3B2026.04 | 56.7 | |
| LMM-R1#Param=3B2026.04 | 56.7 | |
| Perceval#Param=3B2026.04 | 55.8 | |
| R1-VL#Param=7B2026.04 | 55.5 | |
| Vision-R1#Param=7B2026.04 | 55.4 | |
| Qwen2.5-VL + GRPO#Param=3B2026.04 | 55.3 | |
| Perception-R1#Param=3B2026.04 | 54.8 | |
| Jigsaw-R1#Param=3B2026.04 | 54.4 | |
| Qwen2.5-VL#Param=3B2026.04 | 52.1 | |
| R1-VL#Param=2B2026.04 | 38.9 |