Research map / Vision-language and multimodal / Human and embodied modeling
Context-Aware Alignment and Mutual Masking for 3D-Language Pre-Training
Zhao Jin, Munawar Hayat, Yuwei Yang, Yulan Guo, Yinjie Lei
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use CSA and mutual 3D-language masked modeling for pre-training rather than training from scratch (ablation) | 3D-language model · ScanRefer | Acc@0.5: 37.01 -> 39.46 | Improved |
| Add semantic-level alignment (ablation) | 3D-language pre-training · ScanRefer | Acc@0.5: 37.01 -> 38.10 | Improved |
| Add contextual alignment (ablation) | 3D-language pre-training · ScanRefer | Acc@0.5: 37.01 -> 37.63 | Improved |
| Add both semantic-level and contextual alignment (ablation) | 3D-language pre-training · ScanRefer | Acc@0.5: 37.01 -> 38.20 | Improved |
| Add masked language modeling to CSA (ablation) | 3D-language pre-training · ScanRefer | Acc@0.5: 38.20 -> 38.91 | Improved |
| Add masked proposal modeling to CSA (ablation) | 3D-language pre-training · ScanRefer | Acc@0.5: 38.20 -> 39.12 | Improved |
| Use semantic-class reconstruction instead of no MPM reconstruction objective (ablation) | MPM · ScanRefer | Acc@0.5: 38.20 -> 38.52 | Improved |
| Reconstruct raw xyz points instead of using no MPM objective (ablation) | MPM · ScanRefer | Acc@0.5: 38.20 -> 38.18 | No clear effect |
| Reconstruct raw xyz and RGB points instead of using no MPM objective (ablation) | MPM · ScanRefer | Acc@0.5: 38.20 -> 38.35 | No clear effect |
| Restrict proposal-class reconstruction to text-relevant proposals (ablation) | MPM · ScanRefer | Acc@0.5: 38.52 -> 38.85 | Improved |
| Add masked-feature consistency to text-relevant proposal-class reconstruction (ablation) | MPM · ScanRefer | Acc@0.5: 38.85 -> 39.12 | Improved |
| Increase proposal mask ratio from 0% to 75% (ablation) | M3LM · ScanRefer | Acc@0.5: 38.91 -> 39.46 | Improved |
Closest papers by meaning
- Vision-Language Pre-training with Object Contrastive Learning for 3D Scene Understanding (AAAI 2024 · similarity 0.91)
- Beyond 3D VQAs: Injecting 3D Spatial Priors into Vision-Language Models for Enhanced Geometric Reasoning (CVPR 2026 · similarity 0.91)
- VLM-3R: Vision-Language Models Augmented with Instruction-Aligned 3D Reconstruction (CVPR 2026 · similarity 0.90)
- From Flatland to Space: Teaching Vision-Language Models to Perceive and Reason in 3D (NeurIPS 2025 · similarity 0.90)
- Proxy3D: Efficient 3D Representations for Vision-Language Models via Semantic Clustering and Alignment (CVPR 2026 · similarity 0.90)
- MLLMs Need 3D-Aware Representation Supervision for Scene Understanding (NeurIPS 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
