relational-priors

Tag

Cards List
#relational-priors

Relational Priors as Convergence Pressure in LLM-Based Multi-Agent Systems

arXiv cs.CL · 5d ago Cached

This paper studies how making inter-agent relation semantics explicit in LLM-based multi-agent systems acts as convergence pressure, increasing agreement but not reliably improving accuracy. The authors argue relational priors should be used diagnostically and task-specifically, not as a default add-on.

0 favorites 0 likes
← Back to home

Submit Feedback