Research map / Image generation / Text-to-image diffusion
Is This Loss Informative? Faster Text-to-Image Customization by Tracking Objective Dynamics
Anton Voronov, Mikhail Khoroshikh, Artem Babenko, Max Ryabinin
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 |
|---|---|---|---|
| early stopping with DVAR instead of full 6100 iterations | Textual Inversion · 48 concepts | adaptation time (min): 27.0 -> 3.1 | Improved |
| early stopping with DVAR instead of full 6100 iterations | Textual Inversion · 48 concepts | Train CLIP img: 0.840 -> 0.795 | Hurt |
| early stopping with DVAR | Textual Inversion · 48 concepts | Val CLIP txt: 0.209 -> 0.227 | Improved |
| early stopping with DVAR | DreamBooth-LoRA · 48 concepts | adaptation time (min): 7.8 -> 4.9 | Improved |
| early stopping with DVAR | DreamBooth-LoRA · 48 concepts | Val CLIP img: 0.833 -> 0.687 | Hurt |
| early stopping with DVAR | DreamBooth-LoRA · 48 concepts | Train CLIP img: 0.865 -> 0.784 | Hurt |
| early stopping with DVAR | Custom Diffusion · 48 concepts | iterations: 500 -> 348.1 | Improved |
| early stopping with DVAR | Custom Diffusion · 48 concepts | Train CLIP img: 0.755 -> 0.742 | No clear effect |
| replace CLIP-s intermediate sampling with DVAR | Textual Inversion · 48 concepts | adaptation time (min): 9.6 -> 3.1 | Improved |
| replace CLIP-s intermediate sampling with DVAR | Custom Diffusion · 48 concepts | adaptation time (min): 9.7 -> 3.4 | Improved |
| evaluate loss on fixed deterministic batch rather than resampled training loss (ablation) | Textual Inversion · convergence tracking | loss informativeness | Improved |
| resample diffusion timesteps in evaluation batch (ablation) | L_det · convergence tracking | loss informativeness | Hurt |
Closest papers by meaning
- PaRa: Personalizing Text-to-Image Diffusion via Parameter Rank Reduction (ICLR 2025 · similarity 0.90)
- How to Continually Adapt Text-to-Image Diffusion Models for Flexible Customization? (NeurIPS 2024 · similarity 0.89)
- Navigating Text-To-Image Customization: From LyCORIS Fine-Tuning to Model Evaluation (ICLR 2024 · similarity 0.89)
- InstantBooth: Personalized Text-to-Image Generation without Test-Time Finetuning (CVPR 2024 · similarity 0.89)
- Direct Consistency Optimization for Robust Customization of Text-to-Image Diffusion models (NeurIPS 2024 · similarity 0.89)
- Orthogonal Adaptation for Modular Customization of 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.
