Exploring Bad Minima in Transfer Learning

Merge Now, Regret Later: The Hidden Cost of Model Merging Is Adversarial Transferability

arXiv:2509. 23689v2 Announce Type: replace Abstract: Model Merging (MM) has proven to be an effective alternative to multi-task learning, where several fine-tuned models are merged, without access to the tasks' training data, into one model that retains performance across different tasks.

Boosting the Local Invariance for Better Adversarial Transferability

The paper introduces LI-Boost, a technique that enhances the local invariance of adversarial perturbations to improve their transferability across models. By demonstrating a positive correlation between local invariance and transferability, the authors show that LI-Boost significantly boosts five categories of transfer-based attacks on ImageNet, including CNNs, ViTs, defense mechanisms, commercial APIs, and vision-language models. Extensive experiments confirm the method’s effectiveness across diverse architectures and attack types.

Guided Adversarial Robust Transfer Learning with Source Mixing

Guided Adversarial Robust Transfer (GART) learning is a new transfer learning method that relaxes the requirement for source data to closely resemble the target population. By optimizing an adversarial loss over a mixture of source distributions, GART achieves faster convergence and improved prediction performance when target data are scarce. Experiments on simulated data and on multi‑institutional biobank‑linked electronic health records for high‑density lipoprotein cholesterol demonstrate higher robustness and accuracy compared to existing transfer learning approaches.

Transfer of adversarial robustness between perturbation types

Adversarially Trained Linear Transformers Are Optimal Robust In-Context Learners for Gaussian Mixtures

arXiv:2610.07754v1 Announce Type: new Abstract: Adversarial training is one of the most reliable defenses against adversarial attacks, but its high computational cost must generally be paid anew for...

Limits of Transfer Learning

The paper investigates the theoretical limits of transfer learning, demonstrating that careful selection of transferable information and its dependence on target problems is crucial. It establishes that the degree of probabilistic change in a transfer-learning algorithm imposes an upper bound on achievable improvement. These findings extend the algorithmic search framework to a broad class of learning tasks involving transfer.