Reproduce “Trilemma: How Agents choose between Honesty, Confidentiality, and Task Completion” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Johannes Taraz, Abhinav Kumar, Sahar Abdelnabi

The authors introduce TRILEMMA, a simulated Slack staffing task in which agents balance a user’s preference, confidential information, and requests for an honest justification. Across seven open-weight models, 107 of 135 agents pursuing avoidance made false or misleading claims, but only 16 of 70 runs achieved valid staffing. The results describe this scenario, not real-world workplace behavior.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.