UniFAR: A Unified Facet-Aware Retrieval Framework for Scientific Documents

Zheng Dou, Zhao Zhang, Deqing Wang, Yikun Ban, Fuzhen Zhuang · 2026

Read UniFAR: A Unified Facet-Aware Retrieval Framework for Scientific Documents by Zheng Dou, Zhao Zhang, Deqing Wang, Yikun Ban, Fuzhen Zhuang (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-2602-23766.txt.

Description

Zheng Dou and four co-authors present UniFAR, a single system for finding scientific papers in two common ways: from a seed paper, and from a plain question. The 2026 paper notes that past systems handle those two tasks apart, which wastes what each could teach the other.

UniFAR encodes short questions and long papers in one shared space, using several levels of detail and learnable anchors that mark the parts of a paper a question may target. Training joins both tasks, mixing paper-to-paper contrast with question-to-paper contrast and facet-level matching. Tests on three benchmarks, across both tasks and several base models, show steady gains over strong rivals and good transfer across models. The design suits research tools that must serve both browsing and question answering.

Catalog record

CREATORZheng Dou, Zhao Zhang, Deqing Wang, Yikun Ban, Fuzhen Zhuang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE51 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-23766

Browse the full library · About the Tower of Babel