Research map / Image generation / Editing and style transfer
SliderSpace: Decomposing the Visual Capabilities of Diffusion Models
Rohit Gandikota, Zongze Wu, Richard Zhang, David Bau, Eli Shechtman, Nick Kolkin
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 |
|---|---|---|---|
| apply SliderSpace sliders (3 random of 32) to the distilled model | SDXL-DMD · Concept decomposition (6 concepts) | DreamSim inter-image diversity | Improved |
| apply SliderSpace sliders to the distilled model | SDXL-DMD · Concept decomposition (6 concepts) | CLIP score | No clear effect |
| use unsupervised SliderSpace instead of supervised Concept Sliders | SDXL-DMD art styles · Art styles vs reference artist distribution | FID: 32.86 -> 19.12 | Improved |
| use SliderSpace instead of LLM-generated style prompts | SDXL-DMD art styles · Art styles vs reference artist distribution | FID: 38.84 -> 19.12 | Improved |
| use SliderSpace instead of a generic art prompt | SDXL-DMD art styles · Art styles vs reference artist distribution | FID: 76.08 -> 19.12 | Improved |
| use SliderSpace images instead of real artist-name prompts | SDXL-DMD art styles · Art styles user study | win rate (Diverse / Useful %): 54 diverse, 63 useful | Mixed |
| add SliderSpace trained on COCO prompts to distilled model | SDXL-DMD · COCO-30k | FID-30k: 15.52 -> 12.12 | Improved |
| add SliderSpace trained on COCO prompts to distilled model | SDXL-DMD · COCO-30k | CLIP score: 28.92 -> 29.13 | Improved |
| use SliderSpace instead of Concept Sliders for same number of sliders | SliderSpace training · Training time | training speed: 8 times faster | Improved |
Later papers that compared with this method
1 later accepted paper reports SliderSpace as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- RankT2I: A Submodular Framework for Discovering Interpretable and Diverse Semantics in Text-to-Image Models (ECCV 2026 · better on 8 of 8 shared results)
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.
- Concept Sliders — Concept Sliders: LoRA Adaptors for Precise Control in Diffusion Models (ECCV 2024) · better on 1 of 1 shared result
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
- The Hidden Language of Diffusion Models (ICLR 2024 · similarity 0.90)
- Role Bias in Diffusion Models: Diagnosing and Mitigating through Intermediate Decomposition (NeurIPS 2025 · similarity 0.89)
- NoiseCLR: A Contrastive Learning Approach for Unsupervised Discovery of Interpretable Directions in Diffusion Models (CVPR 2024 · similarity 0.88)
- Concept Reachability in Diffusion Models: Beyond Dataset Constraints (ICML 2025 · similarity 0.88)
- I Think, Therefore I Diffuse: Enabling Multimodal In-Context Reasoning in Diffusion Models (ICML 2025 · similarity 0.88)
- CONFORM: Contrast is All You Need for High-Fidelity Text-to-Image Diffusion Models (CVPR 2024 · similarity 0.88)
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.
