Classification on Kandinsky patterns One-red (OR) task (test)
100Accuracy (OR)Gemini 2.5 Pro (GM)
Evaluation results
| Method | Links | |
|---|---|---|
| Gemini 2.5 Pro (GM)Type=LLM2026.04 | 100 | |
| γILP-ViTType=RM2026.04 | 100 | |
| ViTType=VM2026.04 | 80 | |
| YOLO v5 (YL)Type=VM2026.04 | 80 | |
| γILP-VAEType=RM2026.04 | 77 | |
| GPT-o3-mini (G-o3)Type=LLM2026.04 | 63 | |
| GPT-4.1 (G-4.1)Type=LLM2026.04 | 52 | |
| ResNet (CNN)Type=VM2026.04 | 50 | |
| GPT-5 (G-5)Type=LLM2026.04 | 38 | |
| GPT-o4-mini (G-o4)Type=LLM2026.04 | 31 |