General Visual Understanding on AI2D
83.89AccuracyIREASONER
Evaluation Results
| Method | Links | |
|---|---|---|
| IREASONERReward Type=Continuous, Aggregation=Step-level Majority2026.01 | 83.89 | |
| EvoLMM2026.01 | 83.41 | |
| Qwen2.5-VL-7B w/ Discrete Reward + Step-level MajorityReward Type=Discrete, Aggregation=Step-level Majority2026.01 | 82.95 | |
| Vision-Zeroexternal supervision=true2026.01 | 82.64 | |
| Qwen2.5-VL-7B (Baseline)Backbone=Qwen2.5-VL-7B2026.01 | 82.61 | |
| Qwen2.5-VL-7B w/ Discrete RewardReward Type=Discrete2026.01 | 82.18 |