New DASE heuristic optimizes LLM ensemble accuracy by adaptive stopping

Researchers have developed a new heuristic called DASE (Deliberative Adaptive Stopping Ensemble) to improve the accuracy of Large Language Model ensembles. DASE helps ensembles commit to an answer earlier when consensus is reached and uses a fallback for fragmented evidence, preventing degradation from excessive deliberation. The system demonstrated a significant routing gap on the AIME dataset, comparable to existing methods, and showed that adaptive stopping, rather than injection bandwidth, is the primary driver of accuracy gains. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a novel stopping heuristic for LLM ensembles that could improve reasoning accuracy and efficiency.

RANK_REASON This is a research paper detailing a new method for LLM ensembles. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• DASE
• LLM
• AIME
• GPQA

• paper
• model release