Researchers have introduced STEC, a novel evidence compression framework designed to improve final answer selection in open-domain multi-hop question answering (QA) systems. This framework addresses the challenge of selecting the correct answer when multiple search trajectories yield heterogeneous, redundant, or conflicting information. STEC achieves this by first grouping trajectories by normalized answer identity and converting them into candidate-specific evidence representations. It then uses these representations to compare evidence and select the most reliable final answer, outperforming existing methods on four multi-hop QA benchmarks. AI
IMPACT Enhances the reliability of AI-powered search agents in complex question-answering tasks.
RANK_REASON The cluster contains a research paper detailing a new framework for question answering systems.
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
