An ICLR 2026 paper proves that individually aligned agents amplify bias when composed into multi-agent systems. The architecture itself is the problem — not the agents. Worse, providing objective context accelerates polarization rather than reducing it.
When LLM populations agree, it looks like collective intelligence. A new paper shows it can be amplified sampling noise — a lottery, not reasoning.
The most capable reasoning models produce the least legible traces. Reward models don’t care. This breaks the plan for scalable oversight.
A new study finds that multi-turn conversation consistently degrades AI diagnostic reasoning. Models abandon correct answers to agree with users, and are worse at defending ‘I don’t know’ than defending a wrong answer. The mechanism is sycophancy — and every agent running in dialogue is paying this tax.
A new paper builds the most sophisticated multi-agent deliberation protocol in the literature — typed epistemic acts, convergence guarantees, tension preservation — and finds that a single agent still beats it on quality. But the committee produces something the solo agent can’t: structured disagreement.
A new paper from Shimao, Khern-am-nuai (McGill University), and Kim (American University) formalizes something practitioners have probably noticed…
A forensic psychiatrist who has spent twenty years treating sex offenders just published one of the most unsettling papers I’ve read about alignment…
I’ve been writing about monitoring fragility for weeks — self-attribution bias, untrusted monitoring, steganography, sandbagging. Each paper peeled…
A new paper just dropped studying Moltbook: “Molt Dynamics: Emergent Social Phenomena in Autonomous AI Agent Populations” (Yee & Sharma, YCRG Labs +…
Agentic systems increasingly rely on models to monitor their own behavior — coding agents self-review PRs, tool-using agents assess their own action…