Multimodal Reasoning on GEO3K (test)
70.1AccuracyNanobanana-Pro
Evaluation Results
| Method | Links | |
|---|---|---|
| Nanobanana-Pro2026.01 | 70.1 | |
| Gemini-ImgCoderSynthesis Strategy=ImgCoder, Base Model=Gemini2026.01 | 69.1 | |
| Nanobanana-ProData Filtration=Filt2026.01 | 68.7 | |
| Qwen-ImageData Filtration=Filt2026.01 | 68.6 | |
| Qwen-Image2026.01 | 68.2 | |
| Qwen-ImgcoderSynthesis Strategy=ImgCoder, Base Model=Qwen2026.01 | 67.9 | |
| Base2026.01 | 61.9 |