Surgical Video Question Answering on REAL-Colon-VQA (Out-of-template)
44.89BLEU-4SurgLQA
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SurgLQAzero-shot=false2026.05 | 44.89 | 72.98 | 69.59 | 78.93 | — | |
| VideoLLaMA3zero-shot=false2026.05 | 37.08 | 59.82 | 58.55 | 66.97 | — | |
| Qwen3VLzero-shot=false2026.05 | 35.82 | 64.5 | 63.41 | 66.28 | — | |
| RL framework over Digital Twin Representations2026.06 | 35.8 | 57.45 | 58.9 | — | 57.12 | |
| InternVL3 + LoRALLM=-, Modality=Video, Training Paradigm=FT2025.11 | 32.97 | 52.37 | 54.37 | 51.77 | — | |
| SurgViVQALLM=GPT-2, Modality=Video, Training Paradigm=Ours2025.11 | 31.19 | 53.62 | 54.89 | 47.73 | — | |
| SurgViVQAzero-shot=false2026.05 | 31.19 | 53.62 | 54.89 | 47.73 | — | |
| SurgViVQA2026.06 | 31.19 | 53.62 | 54.89 | — | 47.73 | |
| InternVL3zero-shot=false2026.05 | 29.44 | 55.36 | 53.9 | 52.84 | — | |
| MedGemmazero-shot=false2026.05 | 27.16 | 52.8 | 50.43 | 48.62 | — | |
| PitVQALLM=GPT-2, Modality=Image, Training Paradigm=Base2025.11 | 23.63 | 50.03 | 53.22 | 42.93 | — | |
| PitVQAzero-shot=false2026.05 | 23.63 | 50.03 | 53.22 | 42.93 | — | |
| PitVQA2026.06 | 23.63 | 50.03 | 53.22 | — | 42.93 | |
| SurgViVQALLM=Qwen, Modality=Video, Training Paradigm=Ours2025.11 | 16.84 | 37.69 | 35.26 | 38 | — | |
| SurgicalGPTLLM=GPT-2, Modality=Image, Training Paradigm=Base2025.11 | 12.35 | 42.87 | 50.49 | 46.67 | — | |
| SurgicalGPTzero-shot=false2026.05 | 12.35 | 42.87 | 50.49 | 46.67 | — | |
| SurgicalGPT2026.06 | 12.35 | 42.87 | 50.49 | — | 46.67 | |
| Qwen3VLzero-shot=true2026.05 | 6.42 | 36.42 | 41.19 | 47.86 | — | |
| MedGemmaLLM=-, Modality=Image, Training Paradigm=Zero-Shot2025.11 | 5.31 | 30.96 | 27.8 | 34.13 | — | |
| MedGemmaParameters=4B2026.06 | 5.31 | 30.96 | 27.8 | — | 34.13 | |
| InternVL3LLM=-, Modality=Video, Training Paradigm=Zero-Shot2025.11 | 3.87 | 32.1 | 44.34 | 53.73 | — | |
| InternVL3zero-shot=true2026.05 | 3.87 | 32.1 | 44.34 | 53.73 | — | |
| InternVL3Parameters=1B2026.06 | 3.87 | 32.1 | 44.34 | — | 53.73 | |
| Qwen3-VLParameters=8B2026.06 | 3.5 | 42.8 | 46.5 | — | 54.1 | |
| Surgical-LVLM2026.06 | 3.2 | 42 | 45.8 | — | 53.8 | |
| Qwen3-VLParameters=4B2026.06 | 2.1 | 40.5 | 43.8 | — | 52.4 | |
| VideoLLaMA3LLM=-, Modality=Video, Training Paradigm=Zero-Shot2025.11 | 1.78 | 26.17 | 35.15 | 21.47 | — | |
| VideoLLaMA3Parameters=2B2026.06 | 1.78 | 26.17 | 35.15 | — | 21.47 | |
| Qwen2.5LLM=-, Modality=Image, Training Paradigm=Zero-Shot2025.11 | 0.5 | 38.8 | 40.77 | 50.27 | — | |
| Qwen2.5-VLParameters=3B2026.06 | 0.5 | 38.8 | 40.77 | — | 50.27 |