# One Person, One Model, One World: Learning Continual User Representation without Forgetting

arXiv: https://arxiv.org/abs/2009.13724 · HTML: https://mlsanity.com/paper/2009.13724/

## Key facts

- Cited by 30 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 29 Sep 2020.

## Built on by (most cited first, 30 of 30)

- [Recommendation as Language Processing (RLP): A Unified Pretrain, Personalized Prompt & Predict Paradigm (P5)](https://mlsanity.com/paper/2203.13366/) (2022)
- [Large Language Models are Zero-Shot Rankers for Recommender Systems](https://mlsanity.com/paper/2305.08845/) (2023)
- [Online Continual Learning in Image Classification: An Empirical Survey](https://mlsanity.com/paper/2101.10423/) (2021)
- [Chat-REC: Towards Interactive and Explainable LLMs-Augmented Recommender System](https://mlsanity.com/paper/2303.14524/) (2023)
- [Recommendation as Instruction Following: A Large Language Model Empowered Recommendation Approach](https://mlsanity.com/paper/2305.07001/) (2023)
- [Learning Vector-Quantized Item Representation for Transferable Sequential Recommenders](https://mlsanity.com/paper/2210.12316/) (2022)
- [One Model to Serve All: Star Topology Adaptive Recommender for Multi-Domain CTR Prediction](https://mlsanity.com/paper/2101.11427/) (2021)
- [Recent Advances of Foundation Language Models-based Continual Learning: A Survey](https://mlsanity.com/paper/2405.18653/) (2024)
- [A Survey of Large Language Model Empowered Agents for Recommendation and Search: Towards Next-Generation Information Retrieval](https://mlsanity.com/paper/2503.05659/) (2025)
- [A Comprehensive Survey on Trustworthy Recommender Systems](https://mlsanity.com/paper/2209.10117/) (2022)
- [Scaling Law for Recommendation Models: Towards General-purpose User Representations](https://mlsanity.com/paper/2111.11294/) (2021)
- [Continual Learning for Task-oriented Dialogue System with Iterative Network Pruning, Expanding and Masking](https://mlsanity.com/paper/2107.08173/) (2021)
- [Scaling User Modeling: Large-scale Online User Representations for Ads Personalization in Meta](https://mlsanity.com/paper/2311.09544/) (2023)
- [Iterative Network Pruning with Uncertainty Regularization for Lifelong Sentiment Classification](https://mlsanity.com/paper/2106.11197/) (2021)
- [Task Relation-aware Continual User Representation Learning](https://mlsanity.com/paper/2306.01792/) (2023)
- [Async Learned User Embeddings for Ads Delivery Optimization](https://mlsanity.com/paper/2406.05898/) (2024)
- [Universal Item Tokenization for Transferable Generative Recommendation](https://mlsanity.com/paper/2504.04405/) (2025)
- [Scene-adaptive Knowledge Distillation for Sequential Recommendation via Differentiable Architecture Search](https://mlsanity.com/paper/2107.07173/) (2021)
- [MIM: Multi-modal Content Interest Modeling Paradigm for User Behavior Modeling](https://mlsanity.com/paper/2502.00321/) (2025)
- [Generalized User Representations for Transfer Learning](https://mlsanity.com/paper/2403.00584/) (2024)
- [X-Cross: Dynamic Integration of Language Models for Cross-Domain Sequential Recommendation](https://mlsanity.com/paper/2504.20859/) (2025)
- [Continual Low-Rank Adapters for LLM-based Generative Recommender Systems](https://mlsanity.com/paper/2510.25093/) (2025)
- [ERCache: An Efficient and Reliable Caching Framework for Large-Scale User Representations in Meta's Ads System](https://mlsanity.com/paper/2410.06497/) (2024)
- [Dynamic Time-aware Continual User Representation Learning](https://mlsanity.com/paper/2504.16501/) (2025)
- [FOUNDv2: Learning Unified User Quantized Tokenizers for User Representation](https://mlsanity.com/paper/2508.00956/) (2025)
- [Catalog-Native LLM: Speaking Item-ID Dialect with Less Entanglement for Recommendation](https://mlsanity.com/paper/2510.05125/) (2025)
- [SoMeR: Multi-View User Representation Learning for Social Media](https://mlsanity.com/paper/2405.05275/) (2024)
- [Transferable Sequential Recommendation via Vector Quantized Meta Learning](https://mlsanity.com/paper/2411.01785/) (2024)
- [Guarding Digital Privacy: Exploring User Profiling and Security Enhancements](https://mlsanity.com/paper/2504.07107/) (2025)
- [Measuring the stability and plasticity of recommender systems](https://mlsanity.com/paper/2508.03941/) (2025)

## Abstract (arXiv)

Learning user representations is a vital technique toward effective user modeling and personalized recommender systems. Existing approaches often derive an individual set of model parameters for each task by training on separate data. However, the representation of the same user potentially has some commonalities, such as preference and personality, even in different tasks. As such, these separately trained representations could be suboptimal in performance as well as inefficient in terms of parameter sharing. In this paper, we delve on research to continually learn user representations task by task, whereby new tasks are learned while using partial parameters from old ones. A new problem arises since when new tasks are trained, previously learned parameters are very likely to be modified, and as a result, an artificial neural network (ANN)-based model may lose its capacity to serve for well-trained previous tasks forever, this issue is termed catastrophic forgetting. To address this issue, we present Conure the first underlinecontinual, or lifelong, underlineuser underlinerepresentation learner -- i.e., learning new tasks over time without forgetting old ones. Specifically, we propose iteratively removing less important weights of old tasks in a deep user representation model, motivated by the fact that neural network models are usually over-parameterized. In this way, we could learn many tasks with a single model by reusing the important weights, and modifying the less important weights to adapt to new tasks. We conduct extensive experiments on two real-world datasets with nine tasks and show that Conure largely exceeds the standard model that does not purposely preserve such old "knowledge", and performs competitively or sometimes better than models which are trained either individually for each task or simultaneously by merging all task data.