Image Reasoning on WeMath
71.3AccuracyGemini-2.5-Pro
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini-2.5-ProModel Type=Closed-source2026.05 | 71.3 | |
| Claude-4-SonnetModel Type=Closed-source2026.05 | 63 | |
| OPD V→TTraining Strategy=Static On-policy Policy Distillation (Image to Text)2026.04 | 60.95 | |
| MOPDTraining Strategy=MOPD2026.04 | 59.9 | |
| CoPDTraining Strategy=CoPD2026.04 | 59.9 | |
| CoPDTraining Strategy=Co-Evolving Policy Distillation2026.04 | 59.81 | |
| Image-ExpertTraining Strategy=Image-Specific Expert2026.04 | 59.14 | |
| Image-ExpertTraining Strategy=Image-Expert2026.04 | 59.14 | |
| Mixed RLVRTraining Strategy=Mixed RLVR2026.04 | 59.05 | |
| OPD T→VTraining Strategy=Static On-policy Policy Distillation (Text to Image)2026.04 | 58.86 | |
| Mixed RLVRTraining Strategy=Joint optimization on combined budget2026.04 | 57.43 | |
| Text-ExpertTraining Strategy=Text-Specific Expert2026.04 | 55.24 | |
| Text-ExpertTraining Strategy=Text-Expert2026.04 | 55.24 | |
| Video-ExpertTraining Strategy=Video-Expert2026.04 | 53.43 | |
| BaseTraining Strategy=Base Model2026.04 | 52.38 | |
| BaseTraining Strategy=Base2026.04 | 52.38 | |
| GPT-4oModel Type=Closed-source2026.05 | 50.6 | |
| Gemini-2.0-FlashModel Type=Closed-source2026.05 | 47.4 | |
| ATLASLA-GRPOTraining=LA-GRPO2026.05 | 45 | |
| VTS-VModel Type=Agentic Visual2026.05 | 42.8 | |
| LVRModel Type=Latent Visual2026.05 | 41.2 | |
| ATLASGRPOTraining=GRPO2026.05 | 40.3 | |
| MCOTModel Type=Latent Visual2026.05 | 39.6 | |
| BagelModel Type=Unified2026.05 | 39.4 | |
| CoVTModel Type=Latent Visual2026.05 | 38.1 | |
| MonetModel Type=Latent Visual2026.05 | 36.9 | |
| Qwen2.5-VLModel Type=Standard2026.05 | 36.2 | |
| V-ThinkerModel Type=Agentic Visual2026.05 | 32.5 | |
| Gemma-3-27BModel Type=Standard2026.05 | 31.7 | |
| ATLASSFTTraining=SFT2026.05 | 28.9 | |
| Visual CoTModel Type=Agentic Visual2026.05 | 28.6 | |
| AnoleModel Type=Unified2026.05 | 24.7 | |
| LLaVA-OneVision-7BModel Type=Standard2026.05 | 23.1 | |
| MiniGPT-v2Model Type=Standard2026.05 | 11 |