3D Question Answering on OpenEQA
56.2LLM-MatchLLaVA-OV-7B
Evaluation Results
| Method | Links | |
|---|---|---|
| LLaVA-OV-7B2026.03 | 56.2 | |
| GPT4-VInput Frames=50 frames2026.03 | 55.3 | |
| MovieChatBackbone=LLaVA-OV-7B2026.03 | 54.9 | |
| GPT4-VInput Frames=15 frames2026.03 | 54.6 | |
| Claude-3.5 Sonnet2026.03 | 48.7 | |
| Gemini 1.0 Pro Vision2026.03 | 44.9 | |
| LLaMA-2Auxiliary Information=LLaVA-1.5 caption2026.03 | 36.8 | |
| LLaMA-2Auxiliary Information=Sparse Voxel Map2026.03 | 34.3 | |
| Video-ChatGPT2026.03 | 32.1 | |
| AuroraCap2026.03 | 28.9 | |
| LLaMA-2Auxiliary Information=Concept Graph2026.03 | 28.7 | |
| Video-LLaMA2026.03 | 20 |