Adobe's SliderEdit++ Improves Slider-Based Image Editing by 18.5% with New SlideBench
Arman Zarei · hf · 2026-10-26
Adobe Research and University of Maryland present SliderEdit++, addressing unreliable slider control in instruction-based image editing.
Problem and approach:
- Existing sliders produce abrupt visual jumps and non-monotonic changes, and jointly controlled sliders interfere with each other.
- The paper introduces SlideBench, a benchmark for continuous image editing with unified metrics for monotonic progression, perceptual continuity, and inter-slider disentanglement.
- SliderEdit++ regularizes the intermediate editing trajectory (rather than supervising only endpoints) via velocity- and semantic-space regularization plus spectral control, with a disentanglement objective.
Results: 18.5% relative improvement over the strongest baseline overall, while keeping efficiency with a few-step editing backbone.
More from Multimodal
- Real-time fluid sim demo ships 9 looks and 11 physics presets with GPU-baked collision — Michael_Moroz_ · 2026-10-26
- AI-generated frog romance short with a bee punchline delights Reddit — HarshaShastry · 2026-10-26
- 'La dolce vita': egeberkina shares retro Italian-summer AI image set — egeberkina · 2026-10-26
- AI short film teaser 'Kaalakund' shows near-photoreal moody landscapes — KalpnkAIStudio · 2026-10-26
- AI art workflow: The Streets made with Midjourney, GPT Image2 and Seedance 2.5 — Salmaaboukarr · 2026-10-26
- Motion-blur AI portraits show how far image model realism has come — azed_ai · 2026-10-26