https://derekwang2002.github.io/ai-intelligence/en/projects/caught-in-the-act/ Material changes in AI models, agents, engineering and infrastructure, with primary sources and recommendations. en Mon, 02 Nov 2026 16:56:00 GMT https://derekwang2002.github.io/ai-intelligence/en/events/ev-20261008-01/ ev-20261008-01-published Fri, 09 Oct 2026 15:56:33 GMT New TRIAL <p>Caught in the Act scales white-box activation probes to agent-deception monitoring. The authors report 98.8% AUC on SHADE-Arena and detect hidden goals that cannot be inferred from text alone. The FIBS training data, inference code and probe weights are public.</p><p><strong>Why read: </strong>Public data, code and probe weights make agent-deception detection practical to replicate.</p><p>Recommendation: TRIAL</p><p>Sources: <a href="https://arxiv.org/abs/2610.12445">Caught in the Act</a> · <a href="https://github.com/AlignmentResearch/caught-in-the-act-probes">caught-in-the-act-probes</a></p>