Research map / Adversarial robustness and security / Certified and empirical robustness
AutoLoRa: An Automated Robust Fine-Tuning Framework
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Identifies gradient conflict between adversarial and natural objectives in robust fine-tuning (RFT) of feature extractors and proposes AutoLoRa, an automated robust fine-tuning framework.
Problem
Robust fine-tuning (RFT) for adversarial robustness in downstream applications suffers from divergent gradient directions when optimizing adversarial and natural objectives through the feature extractor.
Approach
Observes that jointly optimizing adversarial and natural objectives through the feature extractor yields significantly divergent gradient directions, and builds an automated framework (AutoLoRa) around this finding for robust fine-tuning.
Surprising findings
- Optimizing adversarial and natural objectives through the feature extractor yields significantly divergent gradient directions, which the paper links to the difficulty of RFT.
Experimental setup
Builds on: robust fine-tuning
Closest papers by meaning
- Indirect Gradient Matching for Adversarial Robust Distillation (ICLR 2025 · similarity 0.86)
- Mixture of Adversarial LoRAs: Boosting Robust Generalization in Meta-Tuning (NeurIPS 2024 · similarity 0.86)
- Initialization Matters for Adversarial Transfer Learning (CVPR 2024 · similarity 0.86)
- Mitigating Feature Gap for Adversarial Robustness by Feature Disentanglement (AAAI 2025 · similarity 0.86)
- ROME is Forged in Adversity: Robust Distilled Datasets via Information Bottleneck (ICML 2025 · similarity 0.86)
- Out of Thin Air: Exploring Data-Free Adversarial Robustness Distillation (AAAI 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.