Research map / Vision-language and multimodal / Human and embodied modeling
3DAlign-DAER: Dynamic Attention Policy and Efficient Retrieval Strategy for Fine-grained 3D-Text Alignment at Scale
Yijia Fan, Jusheng Zhang, Kaitong Cai, Jing Yang, Jian Wang, Keze Wang
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 |
|---|---|---|---|
| replace Uni3D-g with 3DAlign-DAER (HAF + MCTS attention policy, pretrained on Align3D-2M) | 3D-text alignment · Objaverse-LVIS zero-shot classification | Top-1 accuracy: 54.2 -> 55.8 | Improved |
| replace Uni3D-g with 3DAlign-DAER | 3D-text alignment · ScanObjectNN zero-shot classification | Top-1 accuracy: 65.2 -> 67.0 | Improved |
| replace SCA3D with 3DAlign-DAER | text-shape retrieval · Text2Shape retrieval S2T | RR@1: 27.22 -> 28.11 | Improved |
| use ERS hierarchical search with 3DAlign-DAER instead of KNN on Uni3D-g | large-scale retrieval · ObjaverseXL 1M text-to-3D retrieval | R@1: 4.1 -> 48.5 | Improved |
| ERS with 3DAlign-DAER instead of the best ANN baseline (DiskANN on Uni3D-g) | large-scale retrieval · ObjaverseXL 1M text-to-3D retrieval | R@1: 36.5 -> 48.5 | Improved |
| 3DAlign-DAER versus ReCon++-L in few-shot linear probing, 1-shot | 3D representations · Objaverse-LVIS few-shot linear probing | Top-1 accuracy: 43.0 -> 47.0 | Improved |
| 3DAlign-DAER versus ReCon++-L in few-shot linear probing, 16-shot | 3D representations · Objaverse-LVIS few-shot linear probing | Top-1 accuracy: 80.0 -> 82.0 | Improved |
Closest papers by meaning
- Escaping Plato's Cave: Towards the Alignment of 3D and Text Latent Spaces (CVPR 2025 · similarity 0.88)
- Riemann-based Multi-scale Attention Reasoning Network for Text-3D Retrieval (AAAI 2025 · similarity 0.87)
- SweetDreamer: Aligning Geometric Priors in 2D diffusion for Consistent Text-to-3D (ICLR 2024 · similarity 0.86)
- A3D: Does Diffusion Dream about 3D Alignment? (ICLR 2025 · similarity 0.85)
- Tango3D: Towards Alignment for Global and Local 2D-3D Correspondence (NeurIPS 2026 · similarity 0.85)
- More Text, Less Point: Towards 3D Data-Efficient Point-Language Understanding (AAAI 2025 · similarity 0.85)
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.
