Part-level Contact MCQ on PARSE-10K (test)
86.2AccuracyOurs
Evaluation results
| Method | Links | |
|---|---|---|
| OursEvaluation protocol=Fine-tuned, Base Model=Qwen3-VL2026.03 | 86.2 | |
| Gemini-2.5-ProEvaluation protocol=Zero-shot2026.03 | 75.6 | |
| GPT-5Evaluation protocol=Zero-shot2026.03 | 75.2 | |
| Claude-Opus-4Evaluation protocol=Zero-shot2026.03 | 73.2 | |
| Qwen3-VLEvaluation protocol=Zero-shot2026.03 | 60.4 | |
| Robobrain2.0Evaluation protocol=Zero-shot2026.03 | 37.2 |