Hi-Q: Hierarchical Evidence-Guided Query Refinement for QA

_reachsumit · x · 2026-09-01

The paper introduces Hi-Q, addressing the granularity mismatch in multi-hop QA.

• Logic: Explicitly checks if retrieved evidence supports the current query unit. Resolved nodes terminate; unresolved nodes are expanded via a dependency-preserving binary operator.
• Structure: Grows a query tree topology determined by corpus support signals rather than fixed templates.
• Performance: In full-corpus retrieval settings, Hi-Q achieves 52.3 EM and 64.0 F1 averaged over three benchmarks, outperforming the iterative retrieval baseline IRCoT by 15.1 EM.

Original post →

More from Research

• Building Non-LLM-as-a-Judge Verification for Context-Scoped AI — One_Stay_1385 · 2026-09-01
• Santa Fe Institute Hiring Postdoc on Thermodynamic Costs of Distributed Computation in Neural Networks and Brains — AnnaCiaunica · 2026-09-01
• Seoul National University Releases MineAmongUs: VLM Agents Learn to Lie in Embodied Social Settings — SeoulNatlUniv · 2026-09-01
• Naver Proposes Verification-Aware Training to Boost Speculative Decoding Draft Models — naver-ai · 2026-09-01
• Research reveals malicious tool descriptions can steal Agent context — askerlee · 2026-09-01
• Tsinghua researchers break 41-year record, prove Dijkstra is not optimal — jedisct1 · 2026-09-01

Research channel →