AI Paper Atlas

Research map / 3D vision / 3D generation

3DrawAgent: Teaching LLM to Draw in 3D with Early Contrastive Experience

Hongcan Xiao, Xinyue Xiao, Yilin Wang, Yue Zhang, Yonggang Qi

CVPR 2026 · 3D vision › 3D generation · approach: Open-vocabulary 3D scene understanding · 2 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 contrastive knowledge extraction (CKE) experience library (ablation)3DrawAgent base LLM · ModelNet40 testCLIP-S: 0.5735 -> 0.6643Improved
use group size K=5 instead of K=2 (ablation)3DrawAgent · ModelNet40 testCLIP-S (Ep 2): 0.6493 -> 0.6643Improved
use group size K=5 instead of K=10 (ablation)3DrawAgent · ModelNet40 testCLIP-S (Ep 2): 0.5612 -> 0.6643Improved
provide ground truth during critique (GT=True) (ablation)3DrawAgent · ModelNet40 testCLIP-S (peak): 0.6643 -> 0.6648No clear effect
provide ground truth during critique (GT=True) (ablation)3DrawAgent · ModelNet40 testCLIP-S (Ep 3): 0.6428 -> 0.6141Hurt
use CLIP-guided pair selection instead of random pairs (ablation)3DrawAgent CKE · ModelNet40 testCLIP-S (Ep 2): 0.5595 -> 0.6643Improved
run more CKE epochs (Ep 2 to Ep 3)3DrawAgent · ModelNet40 testCLIP-S: 0.6643 -> 0.6428Hurt
use Gemini-2.5 Pro instead of DeepSeek-V3.2 as backbone3DrawAgent · Text-to-3D categoryCLIP-S_T: 0.643 -> 0.649Improved
use training-free 3DrawAgent vs trained Diff3DSDiff3DS · Text-to-3D categoryAES: 3.791 -> 4.161Improved
use 3DrawAgent instead of Dream3DVGDream3DVG · Text-to-3D categoryCLIP-S_T: 0.660 -> 0.649No clear effect
use 3DrawAgent instead of Diff3DS in user studyDiff3DS · user preferencepreference rate: 16.67% -> 46.66%Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most 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

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