A Llama-based model trained on serialized user stories unifies item, carousel, and search ranking and outperforms specialist baselines offline while improving some online metrics and reducing latency.
Tran, Jinfeng Rao, Marc Najork, Emma Strubell, and Donald Metzler
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.IR 3years
2026 3verdicts
UNVERDICTED 3roles
background 1polarities
support 1representative citing papers
Presents a taxonomy of generative retrieval failures, empirically identifies issues such as ambiguous docids and low diversity in n-gram methods, and introduces a web-based debugging tool.
A model-agnostic SID alignment update mitigates staleness from temporal drift in user-item interactions for generative retrievers, improving Recall@K and nDCG@K while reducing compute by 8-9x versus full retraining.
citing papers explorer
-
TubiFM: Unified Item, Carousel, and Search Ranking for Streaming Discovery
A Llama-based model trained on serialized user stories unifies item, carousel, and search ranking and outperforms specialist baselines offline while improving some online metrics and reducing latency.
-
Understanding and Debugging Failures in N-Gram-Based Generative Retrieval
Presents a taxonomy of generative retrieval failures, empirically identifies issues such as ambiguous docids and low diversity in n-gram methods, and introduces a web-based debugging tool.
-
Mitigating Collaborative Semantic ID Staleness in Generative Retrieval
A model-agnostic SID alignment update mitigates staleness from temporal drift in user-item interactions for generative retrievers, improving Recall@K and nDCG@K while reducing compute by 8-9x versus full retraining.