BracketRank: Large Language Model Document Ranking via Reasoning-based Competitive Elimination

Abdelrahman Abdallah, Mohammed Ali, Bhawna Piryani, Adam Jatowt · 2026

Read BracketRank: Large Language Model Document Ranking via Reasoning-based Competitive Elimination by Abdelrahman Abdallah, Mohammed Ali, Bhawna Piryani, Adam Jatowt (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-2604-08834.txt.

Description

Abdallah and three colleagues present BracketRank, a setup treating document resorting as a thinking driven contest format. Flexible grouping respects model context limits, and thinking boosted prompts demand step by step fit reasons. A bracket knockout structure with winner and loser tracks enables robust progress plus side by side work. On the BRIGHT thinking benchmark it reaches 26.56 nDCG@10, well ahead of RankGPT-4 and Rank-R1-14B rivals, and it also leads on TREC sets.

Catalog record

CREATORAbdelrahman Abdallah, Mohammed Ali, Bhawna Piryani, Adam Jatowt
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE53366
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.08834

Browse the full library · About the Tower of Babel