AI Paper Atlas

Research map / 3D vision / 3D generation

Scenes as Tokens: Multi-Scale Normal Distributions Transform Tokenizer for General 3D Vision-Language Understanding

Yutao Tang, Cheng Zhao, Gaurav Mittal, Rohith Kukkala, Rama Chellappa, Cheng Peng, Mei Chen

CVPR 2026 · 3D vision › 3D generation · approach: Open-vocabulary 3D scene understanding · 0 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
replace downsampled multi-scale points with multi-scale NDT cells (ablation)NDTokenizer3D · ScanQACiDEr: 94.7 -> 98.6Improved
replace downsampled multi-scale points with multi-scale NDT cells (ablation)NDTokenizer3D · Multi3DRefer valmIoU: 45.3 -> 46.0Improved
use three scales {2,3,4} instead of a single scale (r=3) (ablation)NDTokenizer3D · Multi3DRefer valmIoU: 40.1 -> 46.0Improved
use three scales {2,3,4} instead of a single scale (r=4) (ablation)NDTokenizer3D · SQA3DEM-R: 56.9 -> 57.1Improved
add a fourth, finest scale (r={1,2,3,4}) to the three-scale design (ablation)NDTokenizer3D · ScanQACiDEr: 98.6 -> 94.7Hurt
add a fourth scale (ablation)NDTokenizer3D · SQA3DEM-R: 57.1 -> 57.3No clear effect
increase learnable queries from 100 to 850 (ablation)NDTokenizer3D · ScanQACiDEr: 90.2 -> 98.6Improved
increase queries from 850 to 1200 or 1600 (ablation)NDTokenizer3D · ScanQACiDEr: 98.6 -> 96.3 (1200), 95.9 (1600)Hurt
use NDTokenizer3D instead of 3D-LLaVAgeneralist 3D VLM · Multi3DRefer valmIoU: 42.7 -> 46.0Improved
use NDTokenizer3D instead of 3D-LLaVAgeneralist 3D VLM · SQA3DEM: 54.5 -> 54.4No clear effect
use NDTokenizer3D instead of 3D-LLaVAgeneralist 3D VLM · 3D-POPE (Adversarial)accuracy: 69.88 -> 72.03Improved

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