Who Are They to Each Other? Multi-Agent Reasoning for Speaker Relationship Inference
3.00T1 sourcearXiv cs.MA
Source record
Published by arXiv cs.MA (T1 source). The original is at https://arxiv.org/abs/2609.09628.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryA training-free multi-agent reasoning framework for inferring speaker relationships from spoken conversations, introducing Multi-Role Multi-Agent Debate and Multi-Agent Compete protocols. Evaluated on the Seamless Interaction dataset, results show improvements over baselines in most settings, with human evaluation included.
Why it mattersTwo distinct agent interaction designs (role-based debate and competition-based adjudication) compared on the same task, with human evaluation — a concrete reference for designing multi-agent reasoning pipelines.
Cited by
No citations on record.