๐–ซ๐—‚๐–ข๐–ฐ๐–  \mathsf{LiCQA} : A Lightweight Complex Question Answering System

Sourav Saha, Dwaipayan Roy, Mandar Mitra ยท 2026

Read ๐–ซ๐—‚๐–ข๐–ฐ๐–  \mathsf{LiCQA} : A Lightweight Complex Question Answering System by Sourav Saha, Dwaipayan Roy, Mandar Mitra (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-22182.txt.

Description

Complex questions often have answers scattered across many documents, and most systems for them lean on knowledge graphs or on neural models that are expensive to train. Sourav Saha and two co-authors present LiCQA, a lightweight, unsupervised question-answering system built mainly on corpus evidence. It needs neither a curated graph nor heavy training data to tackle multi-document questions.

In head-to-head tests against two recent systems built on different principles, LiCQA significantly outperforms both on benchmark data while answering with notably lower latency. The result is a system that is both stronger and cheaper to run. For builders of question answering, it shows corpus evidence alone can still carry complex questions.

Catalog record

CREATORSourav Saha, Dwaipayan Roy, Mandar Mitra
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE42 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-22182

Browse the full library ยท About the Tower of Babel