MARQUIS : A Three-Stage Pipeline for Video Retrieval-Augmented Generation Equal Contribution

Debashish Chakraborty, Dengjia Zhang, Jialiang Jin, Hanting Liu, Katherine Guerrerio, Hanxiang Qin, Tyler Skow, Alexander Martin, Reno Kriz, Benjamin Van Durme · 2026

Read MARQUIS : A Three-Stage Pipeline for Video Retrieval-Augmented Generation Equal Contribution by Debashish Chakraborty, Dengjia Zhang, Jialiang Jin, Hanting Liu, Katherine Guerrerio, Hanxiang Qin, Tyler Skow, Alexander Martin, Reno Kriz, Benjamin Van Durme (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-2605-17640.txt.

Description

Chakraborty and colleagues present MARQUIS, a three-stage video retrieval pipeline. Retrieval fails on complex queries one embedding cannot capture, while generation lacks cross-video reasoning and faces memory limits. MARQUIS combines query expansion, fusion and reranking, evidence extraction, and article generation, optionally controlled by a reasoning model. On MAGMaR2026 it lifts retrieval nDCG at 10 from under one fifth to about three quarters. For article generation, ITER-QA-BASE lifts human score from just above 3 to just below 4 over CAG, while MARQUIS-RLM leads non-QA systems on citations.

Catalog record

CREATORDebashish Chakraborty, Dengjia Zhang, Jialiang Jin, Hanting Liu, Katherine Guerrerio, Hanxiang Qin, Tyler Skow, Alexander Martin, Reno Kriz, Benjamin Van Durme
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE89861
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.17640

Browse the full library · About the Tower of Babel