Prompted Identity Degrades Cooperation in Multi-Agent LLM Systems
4.40T1 sourcearXiv cs.MA
Source record
Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2609.35928.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryEmpirical study showing that when multi-agent LLM systems are aware of each other's model family, they form identity-based clusters that impair cooperation. Labeled groups use 30% more rounds and 55% more tokens, with success rates dropping from 96% to 81%; withholding identity labels eliminates the effect.
Why it mattersControlled experiment isolating a previously unmeasured failure mode in mixed-provider multi-agent setups, with a single-line mitigation practitioners can apply to their own systems.
Cited by
No citations on record.
