DAR selects maximally disagreeing agent responses via index-based retention to improve multi-agent debate performance on reasoning benchmarks, especially at scale.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Hear Both Sides: Efficient Multi-Agent Debate via Diversity-Aware Message Retention
DAR selects maximally disagreeing agent responses via index-based retention to improve multi-agent debate performance on reasoning benchmarks, especially at scale.