Question Answering

2 summarised stories about Question Answering, each linking back to the original source. Browse all topics →

Thursday, 16 July 2026

QUBO-Optimized Evidence Selection for Retrieval-Augmented Question Answering with Unconventional Solvers

arXiv cs.CL 18 hours ago

Researchers formulated evidence selection for retrieval-augmented question answering as a Quadratic Unconstrained Binary Optimization (QUBO) problem to identify complementary passages supporting multi-hop answers. The QUBO selector achieved competitive performance with LLM-based selectors on HotpotQA benchmarking. This approach enables RAG systems to use specialized optimization solvers for evidence selection while reserving language models for semantic answer generation, potentially reducing computational costs.

AI Algorithms AI Optimization Question Answering Retrieval-Augmented Generation

LakeQuest: A Three-Domain Benchmark for Grounded Question Answering across Data Lakes

arXiv cs.CL 18 hours ago

Researchers introduced LakeQuest, a benchmark containing 9,846 question-answer pairs designed to evaluate AI systems on retrieving and synthesizing information from realistic, messy data lakes spanning three domains. The benchmark contains questions paired with exact evidence pointers across AI/ML metadata, retail banking, and biomedical drug information datasets. Evaluations of retrieval-augmented generation and agentic systems showed that successful retrieval does not guarantee correct reasoning, with systems failing particularly on relation chaining, policy grounding, and joint tabular question answering.

AI Agents AI Benchmarking AI & Datasets Question Answering