3D Question Answering on SQA3D (EM, EM_R)
60.6Exact Match (EM)GPT4Scene
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GPT4SceneVenue=ICLR 262026.03 | 60.6 | 63.3 | |
| V-LynX-7BEvaluation Protocol=Task-specific models >= 7B, Delta Params=195.0M2026.05 | 60.5 | 62.6 | |
| PAVE-7BEvaluation Protocol=Task-specific models >= 7B, Delta Params=475.0M2026.05 | 59 | 61.4 | |
| LLaVA-OV-7B-FTEvaluation Protocol=Task-specific models >= 7B, Delta Params=161.5M2026.05 | 55.8 | 58.1 | |
| LLaVA-3D-7BEvaluation Protocol=Task-specific models >= 7B, Delta Params=-2026.05 | 55.6 | 57.6 | |
| Chat-SceneVenue=NeurIPS 242026.03 | 54.4 | 57.2 | |
| FastVPruning Ratio=35%2026.04 | 54.22 | 56.77 | |
| Scene-LLM-7BEvaluation Protocol=Task-specific models >= 7B, Delta Params=-2026.05 | 54.2 | — | |
| DVTIEPruning Ratio=35%2026.04 | 54.07 | 56.64 | |
| DVTIEPruning Ratio=65%2026.04 | 54.01 | 56.43 | |
| Baseline (Chat-Scene)Pruning Ratio=0%2026.04 | 53.98 | 56.8 | |
| Fast3DPruning Ratio=35%2026.04 | 53.95 | 56.49 | |
| Fast3DPruning Ratio=65%2026.04 | 53.55 | 56.25 | |
| FastVPruning Ratio=65%2026.04 | 53.27 | 55.91 | |
| DVTIEPruning Ratio=90%2026.04 | 53.09 | 55.55 | |
| Random PruningPruning Ratio=35%2026.04 | 52.74 | 55.28 | |
| ToMePruning Ratio=35%2026.04 | 52.64 | 55.12 | |
| Fast3DPruning Ratio=90%2026.04 | 52.59 | 55.04 | |
| ToMePruning Ratio=65%2026.04 | 52.22 | 54.81 | |
| V-LynX-0.5BEvaluation Protocol=Task-specific models < 7B, Delta Params=68.7M2026.05 | 52.2 | 54.2 | |
| PAVE-0.5BEvaluation Protocol=Task-specific models < 7B, Delta Params=345.9M2026.05 | 51.1 | 52.8 | |
| Random PruningPruning Ratio=65%2026.04 | 51.02 | 53.41 | |
| LEO-7BEvaluation Protocol=Task-specific models >= 7B, Delta Params=-2026.05 | 50 | 52.4 | |
| 3D-LLM-7BEvaluation Protocol=Task-specific models >= 7B, Delta Params=-2026.05 | 49.8 | — | |
| LEOVenue=ICML 242026.03 | 49.4 | 52.2 | |
| Random PruningPruning Ratio=90%2026.04 | 48.48 | 50.99 | |
| FastVPruning Ratio=90%2026.04 | 48.16 | 50.76 | |
| ToMePruning Ratio=90%2026.04 | 47.83 | 50.31 | |
| 3D-LLMVenue=NeurIPS 232026.03 | 47.8 | 49.6 | |
| Chat-3D v2Venue=Arxiv 242026.03 | 45 | 48.1 | |
| LLaVA-OV-0.5B-FTEvaluation Protocol=Task-specific models < 7B, Delta Params=35.2M2026.05 | 44.1 | 45.7 | |
| VideoChat2-7BEvaluation Protocol=Zero-shot Video LLMs, Delta Params=-2026.05 | 37.3 | — | |
| LLaVA-OV-7BEvaluation Protocol=Zero-shot Video LLMs, Delta Params=-2026.05 | 8.3 | — | |
| LLaVA-OV-0.5BEvaluation Protocol=Zero-shot Video LLMs, Delta Params=-2026.05 | 0.8 | 43 |