Research map / Point clouds / Point cloud classification
CloudFixer: Test-Time Adaptation for 3D Point Clouds via Diffusion-Guided Geometric Transformation
Hajin Shim, Changhun Kim, Eunho Yang
This paper tried a common change: Larger batch size
increase batch size from 1 to 64 for CloudFixer-O vs MATE · ModelNet40-C (PointMAE) · average accuracy (%): to 72.68 Improved
What 785 accepted papers report for this change
- 469 improved
- 107 no clear effect
- 48 mixed
- 161 hurt
- Predicting the Emergence of Induction Heads in Language Model Pretraining ICML 2026 · increase batch size (fixed context 1024) · Natural data (CC100) · induction head emergence point (updates): rho=-1.0, p<0.001Improved
- When RL Meets Adaptive Speculative Training: A Unified Training-Serving System ICML 2026 · Increase batch size from 1 to 32 · MiniMax M2.1 FP8 · speedup: 1.57x -> 1.25xHurt
- Untied Ulysses: Memory-Efficient Context Parallelism via Headwise Chunking ICML 2026 · increase per-device batch size to 10 using freed memory · Llama 3-8B, 128K · throughput (TPS): 2464.10 (Ulysses) -> 2526.77 (UPipe)Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 |
|---|---|---|---|
| tailor input adaptation to point clouds (geometric transform with Chamfer objective) instead of 2D-style DDA | DDA · ModelNet40-C (batch size 1) | average accuracy (%): 74.37 -> 78.23 | Improved |
| avoid backpropagation through diffusion model and long generation | DDA · ModelNet40-C | adaptation time per sample (s): 23.6 -> 0.93 | Improved |
| add voting over K=5 transformations (ablation) | CloudFixer · ModelNet40-C (batch size 1) | average accuracy (%): 78.23 -> 79.09 | Improved |
| entropy-minimization TTA (TENT, SHOT) under batch size 1 | Point2Vec · ModelNet40-C (batch size 1) | average accuracy (%): 62.09 -> 4.29 (TENT) | Hurt |
| entropy-minimization and pseudo-label TTA on label-sorted non-i.i.d. stream | Point2Vec · ModelNet40-C (non-i.i.d. label-sorted stream) | average accuracy (%): 60.38 -> 12.80 (TENT) | Hurt |
| add online model adaptation with consistency loss (CloudFixer-O) under batch size 64 | Point2Vec · ModelNet40-C (batch size 64, i.i.d.) | average accuracy (%): 78.85 (TENT) -> 81.49 | Improved |
| increase batch size from 1 to 64 for CloudFixer-O vs MATE (ablation) | CloudFixer-O · ModelNet40-C (PointMAE) | average accuracy (%): to 72.68 | Improved |
| increase MATE update steps (ablation) | MATE · ModelNet40-C (PointMAE) | average accuracy (%): saturates at approximately 61% | No clear effect |
| use small diffusion timestep range t in [0.01T, 0.02T] (ablation) | CloudFixer · ModelNet40-C | average accuracy | Hurt |
| use large diffusion timestep range t in [0.4T, 0.5T] (ablation) | CloudFixer · ModelNet40-C | average accuracy | Hurt |
| replace rotation with affine matrix or remove parameterization (ablation) | CloudFixer · ModelNet40-C | average accuracy | Hurt |
| use squared l2 or diffusion noise-matching loss instead of Chamfer distance (ablation) | CloudFixer · ModelNet40-C | average accuracy | Hurt |
Later papers that compared with this method
1 later accepted paper reports CloudFixer as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- 3D Test-time Adaptation via Graph Spectral Driven Point Shift (ICCV 2025 · better on 6 of 6 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SMART-PC: Skeletal Model Adaptation for Robust Test-Time Training in Point Clouds (ICML 2025 · similarity 0.88)
- Point Cloud Pre-training with Diffusion Models (CVPR 2024 · similarity 0.88)
- Where Precision Meets Efficiency: Transformation Diffusion Model for Point Cloud Registration (AAAI 2025 · similarity 0.88)
- Point-Cache: Test-time Dynamic and Hierarchical Cache for Robust and Generalizable Point Cloud Analysis (CVPR 2025 · similarity 0.88)
- 3DMambaIPF: A State Space Model for Iterative Point Cloud Filtering via Differentiable Rendering (AAAI 2025 · similarity 0.87)
- APML: Adaptive Probabilistic Matching Loss for Robust 3D Point Cloud Reconstruction (NeurIPS 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.
