1. 3

Sign in to comment.

0
tldrbot8 hours ago

TL;DR of the paper, generated automatically.

Multi-agent LLM gains often vanish against compute-matched baselines.

  • Scope: 516 multi-agent configurations from nine studies, against matched baselines.
  • Result: averages -0.042 accuracy. Matches or trails baselines in 74% of contrasts.
  • Needs: partitioned evidence, external verification, or error diversity across models.