Tag
Proposes J-Access, an inference-time audit using the Jacobian lens to measure residual knowledge accessibility in unlearned LLMs, finding that accessibility predicts recovery speed but that directly minimizing it fails to promote genuine deletion.
Introduces Proxy-KD, a novel method for distilling knowledge from black-box large language models (like GPT-4) into smaller models using a proxy model, surpassing both traditional black-box and white-box KD techniques.
This paper introduces idSCD, a white-box method that uses semantic correlation descriptors to identify whether a dataset was used in training a model, outperforming existing baselines across multiple settings.
Describes a white-box memory system for AI agents where every entry is visible and editable, and includes a 'Dream' feature for nighttime memory consolidation and reorganization with one-click rollback.
This paper introduces Semantic Representation Attack (SRA), a novel LLM-agnostic method that optimizes for malicious semantic representations rather than exact text, achieving high attack success rates across multiple open-source models.