Single-Phrase Grounded QA on EndoVis18 VQA standard (test)
76.67AccuracyObject Tokens (Ours)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Object Tokens (Ours)2026.06 | 76.67 | 79.67 | |
| LLaMA-3.2-11BFine-tuned=true2026.06 | 72.59 | 79.78 | |
| EndoChatIn-domain model=true2026.06 | 71.47 | 86.89 | |
| Gemma-3-12BFine-tuned=true2026.06 | 68.58 | 78.46 | |
| GLM-4V-9BFine-tuned=true2026.06 | 45.76 | 51.44 |