The Weight of Teamwork: Does Communication Structure Help LLM Scientists Search for Scientific Laws?

Key points

• Team performance is often reported as one delivered score, even though that score is produced in two stages: a system generates candidate answers and a selection rule returns one of them.1
• Distributing the call budget across more independent trajectories produced the largest measured gain, improving mean held-out fitting score Q by 0.70 at the widest tested allocation.1
• These results support reporting candidate generation and final selection as separate endpoints.1
• They do not identify a communication mechanism or a universally preferred topology; rather, they show that an architecture ranking is not identified by communication structure alone when the rule that returns the final answer is left unspecified.1

Sources (1)

• [1]The Weight of Teamwork: Does Communication Structure Help LLM Scientists Search for Scientific Laws?
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 16, 07:02 PM

  We use symbolic regression as an observable scientific-search setting and compare fixed-budget LLM teams that differ in peer visibility and update timing, evaluating both their frozen candidate archives and the expressions selected from those archives.

  Team performance is often reported as one delivered score, even though that score is produced in two stages: a system generates candidate answers and a selection rule returns one of them.

Extractive summary: sentences quoted from the sources.