Visual Math on ChartQA
89AccuracyOpenVLThinker
Evaluation results
| Method | Links | |
|---|---|---|
| OpenVLThinker#Param=7B2026.04 | 89 | |
| EvoLMM2026.01 | 86.7 | |
| Perceval#Param=3B2026.04 | 86.48 | |
| IREASONERReward Type=Continuous, Aggregation=Step-level Majority2026.01 | 85.78 | |
| VL-Rethinker#Param=7B2026.04 | 85.6 | |
| Qwen2.5-VL-7B w/ Discrete Reward + Step-level MajorityReward Type=Discrete, Aggregation=Step-level Majority2026.01 | 85.42 | |
| VLAA-Thinker#Param=7B2026.04 | 85.36 | |
| Qwen2.5-VL + GRPO#Param=7B2026.04 | 85.16 | |
| LMM-R1#Param=3B2026.04 | 85.04 | |
| Qwen2.5-VL-7B w/ Discrete RewardReward Type=Discrete2026.01 | 84.62 | |
| Jigsaw-R1#Param=3B2026.04 | 84.6 | |
| Perceval#Param=7B2026.04 | 84.44 | |
| Qwen2.5-VL#Param=7B2026.04 | 84.28 | |
| Vision-Zeroexternal supervision=true2026.01 | 84.24 | |
| Qwen2.5-VL-7B (Baseline)Backbone=Qwen2.5-VL-7B2026.01 | 84 | |
| VLM-R1#Param=3B2026.04 | 83.48 | |
| Vision-R1#Param=7B2026.04 | 83.36 | |
| Qwen2.5-VL + GRPO#Param=3B2026.04 | 83.32 | |
| Qwen2.5-VL#Param=3B2026.04 | 83.12 | |
| R1-VL#Param=7B2026.04 | 82.8 | |
| MM-Eureka#Param=7B2026.04 | 82.36 | |
| Perception-R1#Param=3B2026.04 | 81.6 | |
| Pixel-Reasoner#Param=7B, Tool Capability=true2026.04 | 77.36 | |
| DeepEyes#Param=7B, Tool Capability=true2026.04 | 75.84 | |
| R1-VL#Param=2B2026.04 | 68.08 |