Contextual Candour: Role-Aware Graduated Disclosure for LLMs over Mixed-Sensitivity Contexts

Key points

• LLMs increasingly operate over mixed-sensitivity contexts, where they must reason over information ranging from public to confidential while disclosing only what the user is authorised to receive.1
• Existing approaches enforce access control as a binary decision: either fully answering or refusing, without regulating how much or what kind of information is revealed if access is granted, leading to over-refusal or unintended leakage.1
• We model graduated disclosure as a disclosure-aware generation problem and propose a reward-based optimisation framework that shapes responses toward level-appropriate abstractions while penalising unauthorised factual leakage, including paraphrases and approximations.1
• Experiments show that our method substantially reduces information leakage while preserving correctness and utility.1

Sources (1)

• [1]Contextual Candour: Role-Aware Graduated Disclosure for LLMs over Mixed-Sensitivity Contexts
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 11:50 AM

  We introduce graduated disclosure, an access-control formulation in which the level of detail and sensitivity in a model's response is calibrated to the user's authorisation rather than treated as a grant-or-deny decision.

  LLMs increasingly operate over mixed-sensitivity contexts, where they must reason over information ranging from public to confidential while disclosing only what the user is authorised to receive.

Extractive summary: sentences quoted from the sources.