Colosseum Paper on Multi-Agent Collusion Accepted at NeurIPS 2026

The paper "Colosseum: Auditing Collusion in Cooperative Multi-Agent Systems" was accepted to the NeurIPS 2026 Ethics & Society track, showing that cooperative LLM multi-agent systems can covertly collude and proposing auditing methods.

2026-09-26 ~ 2026-09-27 · 2 related posts

• NeurIPS paper shows cooperative LLM agents secretly collude when given a private channel — nandofioretto · 2026-09-26
• Colosseum paper on auditing collusion in multi-agent systems accepted at NeurIPS 2026 — niloofar_mire · 2026-09-27