Research map / Image generation / Editing and style transfer
SliderEdit: Continuous Image Editing with Fine-Grained Instruction Control
Arman Zarei, Samyadeep Basu, Mobina Pournemat, Sayan Nag, Ryan A. Rossi, Soheil Feizi
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 the adapter globally to all image and text tokens (GSTLoRA instead of STLoRA) (ablation) | SliderEdit · single-instruction face edits | Continuity (CLIP): 0.2538 -> 0.2998 | Improved |
| apply the adapter globally to all image and text tokens (GSTLoRA instead of STLoRA) (ablation) | SliderEdit · single-instruction face edits | Disentanglement (DINO): 0.1071 -> 0.1168 | Hurt |
| use STLoRA instead of explicit CFG for edit strength | FLUX-Kontext · single-instruction face edits | Continuity (CLIP): 0.1993 -> 0.2538 | Improved |
| use GSTLoRA instead of Implicit CFG | FLUX-Kontext · single-instruction face edits | Continuity (CLIP): 0.1547 -> 0.2998 | Improved |
| use GSTLoRA instead of Explicit CFG | FLUX-Kontext · single-instruction face edits | Extrapolation (CLIP): 0.2731 -> 0.2648 | Hurt |
| use STLoRA instead of Explicit CFG | FLUX-Kontext · single-instruction face edits | Disentanglement (ID cosine distance): 0.3415 -> 0.2550 | Improved |
| use SliderEdit STLoRA instead of Concept Slider | Concept Slider · single-instruction face edits | Disentanglement (ID cosine distance): 0.7091 -> 0.2550 | Improved |
| use STLoRA instead of Continuous Attribute Control | Continuous Attribute Control · single-instruction face edits | Continuity (CLIP): 0.1891 -> 0.2538 | Improved |
| swap base model from FLUX-Kontext to Qwen-Image-Edit | SliderEdit STLoRA · 3-instruction face edits | Continuity average: 0.3691 -> 0.4345 | Improved |
| swap base model from FLUX-Kontext to Qwen-Image-Edit | SliderEdit STLoRA · 1-instruction face edits | Continuity average: 0.2709 -> 0.2319 | Hurt |
| swap base model from FLUX-Kontext to Qwen-Image-Edit | SliderEdit STLoRA · 1-instruction face edits | Extrapolation average: 0.2630 -> 0.3214 | Improved |
| swap base model from FLUX-Kontext to Qwen-Image-Edit | SliderEdit STLoRA · 3-instruction face edits | Disentanglement average (lower better): 0.2762 -> 0.3630 | Hurt |
Closest papers by meaning
- InsightEdit: Towards Better Instruction Following for Image Editing (CVPR 2025 · similarity 0.91)
- AnyEdit: Mastering Unified High-Quality Image Editing for Any Idea (CVPR 2025 · similarity 0.90)
- UltraEdit: Instruction-based Fine-Grained Image Editing at Scale (NeurIPS 2024 · similarity 0.90)
- Balancing Preservation and Modification: A Region and Semantic Aware Metric for Instruction-Based Image Editing (ICML 2025 · similarity 0.89)
- SmartEdit: Exploring Complex Instruction-based Image Editing with Multimodal Large Language Models (CVPR 2024 · similarity 0.89)
- FireEdit: Fine-grained Instruction-based Image Editing via Region-aware Vision Language Model (CVPR 2025 · similarity 0.89)
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.
