Cube Three-View Reasoning
55AccuracyGemini-2.5-Flash
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini-2.5-FlashModel Type=Commercial2026.06 | 55 | |
| Gemini-2.5-ProModel Type=Commercial2026.06 | 55 | |
| InternVL3.5-8BModel Type=Open-Source2026.06 | 51 | |
| Qwen3-32BModel Type=Open-Source2026.06 | 48.1 | |
| InternVL3.5-14BModel Type=Open-Source2026.06 | 46.6 | |
| Mini-o3Model Type=Tool-Augmented2026.06 | 39.7 | |
| w/o RLModel Type=Ablation2026.06 | 39 | |
| Qwen3-235BModel Type=Open-Source2026.06 | 38.5 | |
| Qwen3-30BModel Type=Open-Source2026.06 | 37.7 | |
| PERIA-8BModel Type=Proposed2026.06 | 36.3 | |
| VTool-R1-7BModel Type=Tool-Augmented2026.06 | 35.6 | |
| R1-OneVision-7BModel Type=Tool-Augmented2026.06 | 33.7 | |
| GPT-5Model Type=Commercial2026.06 | 32.9 | |
| w/o ToolModel Type=Ablation2026.06 | 30.4 | |
| w/o SFTModel Type=Ablation2026.06 | 29.1 | |
| Qwen3-8BModel Type=Backbone Baseline2026.06 | 21.2 |