AI Paper Atlas

Research map / Point clouds / Point cloud classification

PointAlign: Feature-Level Alignment Regularization for 3D Vision-Language Models

Yuanhao Su, Shaofeng Zhang, Xiaosong Jia, Qi Fan

CVPR 2026 · Point clouds › Point cloud classification · approach: Point-language and open-vocabulary 3D · 1 citations (OpenAlex)

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 changedWhereMeasuredEffect
add feature-level alignment loss at LLM layer 16 (lambda 0.1 vs 0) (ablation)MiniGPT-3D Stage 2 · ModelNet40+Objaverse classificationaverage accuracy (%): 64.72 -> 66.08Improved
add PointAlign over MiniGPT-3D on Objaverse classificationMiniGPT-3D · Objaverse classification (instruction prompt)accuracy (%): 65.00 -> 72.50Improved
add PointAlign over MiniGPT-3D on ModelNet40MiniGPT-3D · ModelNet40 classification (average)accuracy (%): 61.24 -> 61.17No clear effect
add PointAlign to MiniGPT-3D for captioningMiniGPT-3D · Objaverse captioningQwen2-72B-Instruct score: 48.17 -> 53.05Improved
add PointAlign to MiniGPT-3D for captioningMiniGPT-3D · Objaverse captioningSentence-BERT / SimCSE: S-BERT 49.54 -> 49.94; SimCSE 51.39 -> 51.32No clear effect
use cosine loss instead of L1 or L2 for alignment (ablation)PointAlign · classification averageaccuracy (%): L1 65.72 / L2 65.57 -> cosine 66.08Improved
align at layer 16 rather than layers 8-32 or multi-layer (ablation)PointAlign · classification averageaccuracy (%): layer 12: 65.14, layer 28: 64.81 -> layer 16: 66.08Improved
increase alignment weight lambda from 0.1 to 0.3-0.9 (ablation)PointAlign · classification averageaccuracy (%): 66.08 -> 64.39 to 65.08Hurt
use 3 linear layers in the alignment projector instead of 1, 2 or 4 (ablation)PointAlign · classification averageaccuracy (%): 1 layer 64.45; 2 layers 65.48; 4 layers 65.24 -> 3 layers 66.08Improved
use Q-Former output as alignment target instead of projector features (ablation)PointAlign · classification averageaccuracy (%): 65.09 (Projector-mid), 64.91 (Projector) -> 66.08Improved
add alignment regularization when training data fraction grows from 50% to 100% (ablation)MiniGPT-3D captioning · Objaverse captioning (Qwen2 score)score: baseline 52.72 -> 48.17 (100%); aligned 53.05 at 100%Improved
add alignment with only 10% of training data (ablation)MiniGPT-3D captioning · Objaverse captioning (10% data)Qwen2 score: 41.77 -> 41.95No clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy