Toward Trustworthy AI: Multi-Target Adversarial Attacks and Robust Defenses for Continuous Data Summarization

In three linesStudy of adversarial attacks on continuous data summarization via DR-submodular optimization. Authors formulate multi-target attacks as min-max problem and propose robust defenses through regularized max-min. Approximation algorithms with theoretical guarantees validated on real benchmarks.

Summary generated by Claude — human-verified