| Paper | Code | CIDEr | ROUGE | ModelName | ReleaseDate |
|---|---|---|---|---|---|
| GAP-MLLM: Geometry-Aligned Pre-training for Activating 3D Spatial Perception in Multimodal Large Language Models | ✓ Link | 84.7 | 63.1 | GAP-MLLM-3B | 2026-03-17 |
| 3D-Aware VLMs with Implicit and Explicit Geometries | ✓ Link | 80.4 | VLM-IE3D | 2026-07-23 | |
| 3D-LLaVA: Towards Generalist 3D LMMs with Omni Superpoint Transformer | ✓ Link | 78.8 | 57.7 | 3D-LLaVA | 2025-01-02 |
| Merge3D: Efficient 3D Multimodal LLMs via Joint 2D-3D Token Merging | 73.4 | 61.6 | Merge3D (30% token retention) | 2026-06-03 | |
| Imagine3D-LLM: Teaching MLLMs to Imagine 3D Scenes Before Answering | ✓ Link | 67.6 | Imagine3D-LLM | 2026-09-29 |