1 paper · 1 filter
Tommaso Giovannelli, Griffin D. Kent
Multi-agent debate has emerged as a promising approach for improving LLM reasoning on ground-truth tasks, yet current methodologies face certain structural limitations: debate tend…