AI Agents Spontaneously Form In-Groups by Model Family, Study Finds

A Sapienza University study finds that LLM agents in multi-agent systems spontaneously form cliques by model family once identities are revealed, even when the task offers no incentive; hiding model names shifts success rates from 96% down and back up.

2026-10-08 ~ 2026-10-08 · 2 related posts

• Hide Model Names From Agents: Labels Cost 55% More Tokens and Drop Success to 81% — alex_verem · 2026-10-08
• LLM Agents Form Factions by Model Family, Costing 30% More Rounds and 55% More Tokens — alex_verem · 2026-10-08