KuaFu is a unified behavior-compression layer whose minimal unit is one behavior item, with fidelity-oriented four-stage training and layered intermediate evaluation, which matches or exceeds uncompressed single-task production models on all five headline metrics, raises per-GPU throughput by 37%-350%, and saves 190 GPUs.
Abstract
Learning user preferences from long behavioral sequences is fundamental to large-scale advertising, recommender systems, and conversational agents. In industrial systems, preference modeling is typically task-wise, with different downstream tasks independently processing the same user history to capture different aspects of preference. This repeated processing creates substantial redundancy at billion-user scale, especially as preferences evolve and representations require frequent updates. We present KuaFu, a unified user preference modeling system that replaces repeated user-level sequence encoding with reusable item-level compressed representations. Each behavioral item is encoded once and shared across users and downstream tasks, while a user's history is built chronologically and extended with new item representations as they arrive, letting KuaFu capture evolving preferences without re-encoding the entire history. To reduce inference and storage overhead, KuaFu introduces a two-axis projector: token-axis compression shortens behavior sequences by up to 15x and reduces each item from 150 KB to 10 KB, while width-axis compression reduces the cached representation to 0.5 KB. We further develop a four-stage training pipeline consisting of reconstruction pre-training, preference-understanding post-training, joint compression-generation training, and hallucination-aware reinforcement learning, progressively improving representation fidelity and downstream utility. Across six MRQA and five RecBench benchmarks, KuaFu consistently outperforms existing compression methods at comparable ratios, with gains of up to 17.7 EM and 1.90 AUC points respectively. KuaFu has been deployed and evaluated on the Tencent advertising and recommendation platform for ten months. Over this period, it improves per-GPU throughput by up to 350%, saves 190 GPUs, and increases overall GMV by 1.37%.
The contribution is a system-level integration that makes long-term, multi-interest, and multimodal modeling jointly deployable in a real-time production pipeline, together with the engineering practices required to sustain it.
Yong-Kang Fu, Bei-Ning Bao, Yu Jiang et al.· 0 citations
Inspired by the recent success of Byte Latent Transformer, DP-Rec is proposed, a dynamic latent patching architecture for recommendation that scales effectively to long sequences and achieves a superior efficiency–accuracy trade-off over both non-compressed and fixed-size compression baselines.
Dwipam Katariya, T. Caputo, Akshat Shreemali et al.· Proceedings of the 20th ACM...· 0 citations
Experiments conducted on movie, book and electronics benchmark datasets demonstrate that GradSup outperforms iterative fine-tuning to provide scalable and personalised recommendation that is consistently sustained above the frozen LLM backbone.
Kanishka Dandeniya, C. Dasanayaka, Daswin de Silva et al.· Proceedings of the 20th ACM...· 0 citations
ChronicleRec is a pre-train-and-transfer framework that compresses an ultra-long behavior sequence once into a chronologically ordered set of Chronicle Tokens, which can be cached per user, decoupling ultra-long sequence modeling from online candidate scoring.
Chengkai Huang, Yu-Bin Sheng, Liang Guo et al.· 0 citations
RDSR, a Rate–Distortion–based Sequential Recommendation framework grounded in a task-oriented rate-utility view, combines fixed-capacity token selection with VIB-based latent compression to retain task-relevant information under explicit rate control, effectively reducing sequence length and computational overhead.
Woo-Seung Kang, Minje Kim, Suwon Lee et al.· Proceedings of the 20th ACM...· 0 citations
In-context learning (ICL) can be amortized into latent objects (task vectors, function vectors, context vectors) that recover few-shot behavior at zero-shot inference cost, but recent theory shows a static vector acts as a single synthetic demonstration and must fail on high-rank mappings such as word-level bijections....
Gunmay Jhingran· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduAug 27, 2026
A new machine-learning framework aims to improve the success rate of computational protein design while moving away from results that reproduce sequences found in nature.