Visual Relation MCQ on PARSE-10K (test)
97.4AccuracyOurs
Evaluation results
| Method | Links | |
|---|---|---|
| OursEvaluation protocol=Fine-tuned, Base Model=Qwen3-VL2026.03 | 97.4 | |
| Qwen3-VLEvaluation protocol=Zero-shot2026.03 | 86.2 | |
| Gemini-2.5-ProEvaluation protocol=Zero-shot2026.03 | 85 | |
| GPT-5Evaluation protocol=Zero-shot2026.03 | 82.1 | |
| Claude-Opus-4Evaluation protocol=Zero-shot2026.03 | 80.3 | |
| Robobrain2.0Evaluation protocol=Zero-shot2026.03 | 60.8 |