Research map / Architectures and efficiency / Pruning and efficient tuning
PEFTDiff: Diffusion-Guided Transferability Estimation for Parameter-Efficient Fine-Tuning
Prafful Kumar Khoba, Zijian Wang, Chetan Arora, Mahsa Baktashmotlagh
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 |
|---|---|---|---|
| combine intra-class compactness and inter-class separability scores (ablation) | PEFTDiff · VTAB-1k | weighted Kendall tau_w: 0.174 / 0.22 -> 0.517 | Improved |
| use diffusion step t=10 instead of t=5 or t=20 (ablation) | PEFTDiff · VTAB-1k | weighted Kendall tau_w: 0.439 / 0.107 -> 0.517 | Improved |
| use kNN graph with k=20 (ablation) | PEFTDiff · VTAB-1k | weighted Kendall tau_w | Improved |
| use diffusion-based score instead of LogME for PEFT ranking | PEFT selection · VTAB-1k | weighted Kendall tau_w: 0.306 -> 0.517 | Improved |
| use diffusion-based score on structured VTAB tasks | PEFTDiff · VTAB-1k Structured | weighted Kendall tau_w: 0.397 vs LogME 0.391 | Mixed |
| use diffusion-based score on FGVC instead of best baseline | PEFT selection · FGVC (5 datasets) | weighted Kendall tau_w: 0.311 -> 0.462 | Improved |
| use diffusion score instead of exhaustive PEFT fine-tuning for selection | PEFT selection · VTAB-1k | computation time (min): 11074.5 -> 55.44 | Improved |
Closest papers by meaning
- Transfer Learning for Diffusion Models (NeurIPS 2024 · similarity 0.87)
- Diffusion Tuning: Transferring Diffusion Models via Chain of Forgetting (NeurIPS 2024 · similarity 0.87)
- Domain Guidance: A Simple Transfer Approach for a Pre-trained Diffusion Model (ICLR 2025 · similarity 0.86)
- Provable Sample-Efficient Transfer Learning Conditional Diffusion Models via Representation Learning (NeurIPS 2025 · similarity 0.86)
- SVFT: Parameter-Efficient Fine-Tuning with Singular Vectors (NeurIPS 2024 · similarity 0.86)
- Parameter Efficient Fine-tuning via Cross Block Orchestration for Segment Anything Model (CVPR 2024 · similarity 0.86)
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.
