Papers for

automated evaluation designers

Papers whose findings have a practical use for this group, as judged from the abstract. Open a paper to read what it means in practice.

LLM systems face risks of groupthink and feedback loops

LLM Multi-Agent Systems Are Vulnerable to Herding and Runaway Cascades

Abstract: Large language models (LLMs) increasingly operate in multi-agent systems where they collaborate by responding to others' choices, creating endogenous feedback loops with vulnerability to social influence, herding behavior and runaway cascades. Following the design of the classic Music Lab experiment, we tested this vulnerability in large populations of LLM agents that recommended unfamiliar songs, with or without exposure to cumulative recommendation counts. Exposure led to runaway cascades that generated extreme and unpredictable collective recommendations increasingly divergent from a song's intrinsic appeal. Compared to humans, LLM populations were generally far more susceptible to herding behavior, and unlike humans, LLM cascades were less able to self-correct. This vulnerability has implications for multi-agent collaboration in recommender systems, automated evaluation, complex problem-solving, and unauthorized conspiracies.

Thu 22 OctArtificial IntelligenceMultiagent Systems

The gist

Large language models (LLMs) working together can end up influencing each other too much, like people in a crowd copying others without thinking. The authors show that when many LLM agents recommend songs and see what others have chosen, they tend to follow the crowd, creating extreme and unpredictable popularity patterns that don’t match the actual quality of the songs. This effect is stronger and less self-correcting than what happens with humans. These findings matter for how AI systems work together in choosing recommendations, solving problems, or making decisions.

Open → 2610.30769v1