Research map / 3D vision / 3D generation
JointDreamer: Ensuring Geometry Consistency and Text Congruence in Text-to-3D Generation via Joint Score Distillation
ChenHan Jiang, Yihan Zeng, Tianyang Hu, Songcen Xu, Wei Zhang, Hang Xu, Dit-Yan Yeung
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 |
|---|---|---|---|
| replace SDS with Joint Score Distillation (energy-based inter-view coherence) (ablation) | text-to-3D with SD-V2.1 · MS-COCO 30% subset | CLIP Score: 20.0 -> 27.6 | Improved |
| replace SDS with JSD (ablation) | text-to-3D with SD-V2.1 · MS-COCO 30% subset | FID: 429.2 -> 360.7 | Improved |
| add CFG Switching (CFG 30 then 50 after 5K iterations) (ablation) | JointDreamer · MS-COCO 30% subset | CLIP Score: 27.6 -> 28.2 | Improved |
| add Geometry Fading (reduce density lr after 5K iterations) (ablation) | JointDreamer · MS-COCO 30% subset | CLIP Score: 28.2 -> 28.8 | Improved |
| use JSD with multi-view synthesis energy vs SDS (ablation) | text-to-3D · 16 Janus prompts | Janus rate: 100% -> 6.2% | Improved |
| use image-to-image translation model as energy function instead of MVS (ablation) | JSD · 16 Janus prompts | Janus rate: 6.2% -> 31.2% | Hurt |
| JSD with MVS energy costs relative to SDS (ablation) | JSD · 16 Janus prompts | Train time (min): 50 -> 54 min; 16.1 -> 19.4 G | No clear effect |
| replace gradient of coherence energy with random values (ablation) | JSD with classification model · qualitative | shape quality | Hurt |
| naive SDS + MVDream combination instead of JSD (ablation) | text-to-3D · qualitative | text congruence / Janus | Hurt |
| JointDreamer vs MVDream | text-to-3D · MS-COCO object-centric subset | CLIP R-Precision: 33.6 -> 88.5 | Improved |
Later papers that compared with this method
1 later accepted paper reports JointDreamer as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- CoherenDream: Boosting Holistic Text Coherence in 3D Generation via Multimodal Large Language Models Feedback (AAAI 2026 · better on 3 of 3 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
- Let 2D Diffusion Model Know 3D-Consistency for Robust Text-to-3D Generation (ICLR 2024 · similarity 0.92)
- Retrieval-Augmented Score Distillation for Text-to-3D Generation (ICML 2024 · similarity 0.91)
- LucidDreamer: Towards High-Fidelity Text-to-3D Generation via Interval Score Matching (CVPR 2024 · similarity 0.91)
- Consistent3D: Towards Consistent High-Fidelity Text-to-3D Generation with Deterministic Sampling Prior (CVPR 2024 · similarity 0.91)
- Taming Mode Collapse in Score Distillation for Text-to-3D Generation (CVPR 2024 · similarity 0.90)
- IM-3D: Iterative Multiview Diffusion and Reconstruction for High-Quality 3D Generation (ICML 2024 · similarity 0.90)
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.
