Research map / Point clouds / Point cloud classification
Point Ladder Tuning: Parameter-Efficient Hierarchical Adaptation for 3D Point Cloud Understanding
Junlin Chang, Longhao Zou, Rui Li
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 |
|---|---|---|---|
| add LGF fusion between HLN and backbone (ablation) | HLN · ScanObjectNN PB_T50_RS | overall accuracy: 80.74 -> 83.34 | Improved |
| add dynamic prompt (ablation) | HLN + LGF · ScanObjectNN PB_T50_RS | overall accuracy: 83.34 -> 84.52 | Improved |
| add SSF scale-and-shift tuning (ablation) | HLN + LGF + dynamic prompt · ScanObjectNN PB_T50_RS | overall accuracy: 84.52 -> 85.53 | Improved |
| add dynamic prompt to full PLT (ablation) | PLT (with SSF) · ScanObjectNN PB_T50_RS | overall accuracy: 84.66 -> 85.53 | Improved |
| use LGF with softmax instead of additive fusion (ablation) | PLT · ScanObjectNN PB_T50_RS | overall accuracy: 85.01 -> 85.53 | Improved |
| use LGF with softmax instead of concatenation (ablation) | PLT · ScanObjectNN PB_T50_RS | overall accuracy: 84.63 -> 85.53 | Improved |
| use LGF with softmax instead of sigmoid (ablation) | PLT · ScanObjectNN PB_T50_RS | overall accuracy: 84.59 -> 85.53 | Improved |
| fuse local and global features instead of local only (ablation) | PLT · ScanObjectNN PB_T50_RS | overall accuracy: 82.86 -> 85.53 | Improved |
| use raw-point HLN instead of token-center HLN (ablation) | PLT · ScanObjectNN PB_T50_RS | overall accuracy: 84.73 -> 85.53 | Improved |
| use raw-point HLN instead of token-center HLN with standard head (ablation) | PLT · S3DIS | mIoU: 59.0 -> 60.5 | Improved |
| use lightweight segmentation head instead of standard head (ablation) | PLT · S3DIS | mIoU: 60.5 -> 61.5 | Improved |
| use raw-point HLN instead of random centers (ablation) | PLT · ScanObjectNN PB_T50_RS | overall accuracy: 85.01 -> 85.53 | Improved |
Closest papers by meaning
- Adaptive Point-Prompt Tuning: Fine-Tuning Heterogeneous Foundation Models for 3D Point Cloud Analysis (TPAMI 2026 · similarity 0.91)
- Point-PEFT: Parameter-Efficient Fine-Tuning for 3D Pre-trained Models (AAAI 2024 · similarity 0.90)
- On Geometry-Enhanced Parameter-Efficient Fine-Tuning for 3D Scene Segmentation (NeurIPS 2025 · similarity 0.89)
- GAPrompt: Geometry-Aware Point Cloud Prompt for 3D Vision Model (ICML 2025 · similarity 0.87)
- PointLoRA: Low-Rank Adaptation with Token Selection for Point Cloud Learning (CVPR 2025 · similarity 0.87)
- Sonata: Self-Supervised Learning of Reliable Point Representations (CVPR 2025 · similarity 0.87)
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.
