Pith. sign in

REVIEW 9 cited by

KuaiFormer: Transformer-Based Retrieval at Kuaishou

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2411.10057 v1 pith:6XVUN5UG submitted 2024-11-15 cs.IR cs.AIcs.LG

Chi Liu , Jiangxia Cao , Rui Huang , Kai Zheng , Qiang Luo , Kun Gai , Guorui Zhou This is my paper · ORCID
classification cs.IRcs.AIcs.LG
keywords retrievalrecommendationkuaiformerlarge-scalebeencontentkuaishousystems
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

In large-scale content recommendation systems, retrieval serves as the initial stage in the pipeline, responsible for selecting thousands of candidate items from billions of options to pass on to ranking modules. Traditionally, the dominant retrieval method has been Embedding-Based Retrieval (EBR) using a Deep Neural Network (DNN) dual-tower structure. However, applying transformer in retrieval tasks has been the focus of recent research, though real-world industrial deployment still presents significant challenges. In this paper, we introduce KuaiFormer, a novel transformer-based retrieval framework deployed in a large-scale content recommendation system. KuaiFormer fundamentally redefines the retrieval process by shifting from conventional score estimation tasks (such as click-through rate estimate) to a transformer-driven Next Action Prediction paradigm. This shift enables more effective real-time interest acquisition and multi-interest extraction, significantly enhancing retrieval performance. KuaiFormer has been successfully integrated into Kuaishou App's short-video recommendation system since May 2024, serving over 400 million daily active users and resulting in a marked increase in average daily usage time of Kuaishou users. We provide insights into both the technical and business aspects of deploying transformer in large-scale recommendation systems, addressing practical challenges encountered during industrial implementation. Our findings offer valuable guidance for engineers and researchers aiming to leverage transformer models to optimize large-scale content recommendation systems.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 9 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. WatchLens: A Configurable Platform for Online Video Recommendation Experiments

    cs.IR 2026-08 conditional novelty 6.0 of 10

    WatchLens is an open-source platform that attaches exposure context to every playback event, enabling controlled online experiments on how recommendation policies shape viewing behavior.

  2. Embedding Items at Scale: Comparing GNN-Based and ID-Based Item Embeddings in the Yandex Ecosystem

    cs.IR 2026-07 conditional novelty 6.0 of 10

    Pretrained GNN item embeddings outperform end-to-end ID embeddings on a small dataset, but not in two large-scale Yandex production recommender systems.

  3. Real-Time Hard Negative Sampling via LLM-based Clustering for Large-Scale Two-Tower Retrieval

    cs.IR 2026-07 unverdicted novelty 6.0 of 10

    Cluster-based real-time out-of-batch negatives drawn from LLM media embeddings outperform industry-standard negative sampling for two-tower retrieval and cut popularity bias.

  4. Request-Only Optimization for Recommendation Systems

    cs.IR 2025-07 conditional novelty 6.0 of 10

    A request-level training data format eliminates duplicate user features, increasing storage efficiency and training throughput while enabling larger recommendation architectures.

  5. Unifying Generative Recall and Multi-Objective Ranking in a Single Decoder-Only Sequence

    cs.IR 2026-07 conditional novelty 5.5 of 10

    A single decoder-only sequence with dual-query prefix-causal attention and ranking-side LoRA unifies generative SID recall and multi-objective ranking, with offline and online gains at Kuaishou.

  6. MISS: Multi-Modal Tree Indexing and Searching with Lifelong Sequential Behavior for Retrieval Recommendation

    cs.IR 2025-08 conditional novelty 5.0 of 10

    MISS builds a k-means index tree on interaction-supervised multi-modal embeddings and adds two behavior search units (Co-GSU, MM-GSU) plus ESU/MMoE, reporting ~30-47% relative recall gains over TDM+MMoE on Kuaishou da...

  7. Correcting the LogQ Correction: Revisiting Sampled Softmax for Large-Scale Retrieval

    cs.IR 2025-07 conditional novelty 5.0 of 10

    The paper derives and tests a refined logQ-correction loss that drops the positive item from the sampled denominator and scales each positive example by one minus its estimated model probability.

  8. Synergizing Implicit and Explicit User Interests: A Multi-Embedding Retrieval Framework at Pinterest

    cs.IR 2025-06 conditional novelty 5.0 of 10

    A production recommender framework combines a differentiable clustering module for implicit interests and conditional retrieval for explicit followed topics, deployed at Pinterest home feed.

  9. GR-LLMs: Recent Advances in Generative Recommendation Based on Large Language Models

    cs.IR 2025-07 unverdicted novelty 3.0 of 10

    A survey of LLM-based generative recommendation systems, covering application settings, training pipelines, industrial deployment challenges, and future directions.

Pith tools