Get Started
Home
Topics
Search
Library
Research questionHow can language models personalize for many users without costly per-user model storage?Keeping a separate model for every user quickly becomes expensive. The challenge is retaining individual preferences while keeping per-user storage manageable.
AI
AI Memory
Evaluation & Benchmarks
Inference Optimization
LLM Pretraining & Post-training
Machine Learning
Natural Language Processing
Research Paper
Latest papersRecent research connected to this question, newest first.Aplaud: Adaptive Personalized Low-Rank Decomposition for User-Specific LLMThe source evaluates a low-rank personalization framework for fine-tuned LLMs across users in survey-response prediction. Its evidence concerns the reported personalization, generalization, and inference-efficiency experiments; it does not specify additional access requirements or establish applicability beyond this task.research paper · Sep 4, 2026PLUME: Parameter-Efficient Personalization of Large Language Models via Low-Rank User Modulation in Shared SubspacesThe source uses a shared task-specific subspace with lightweight user-specific modulation and evaluates it for personalized text generation. Evidence is limited to the reported personalization tasks and model-training settings; it does not establish deployment requirements beyond those settings.research paper · Sep 4, 2026Beyond Retrieval: Learning Compact User Representations for Scalable LLM PersonalizationThe source studies a prefix-based personalization framework for LLMs and reports results on six LaMP tasks spanning classification, rating, and generation. Its evidence compares prompt-based and model-based baselines and includes parameter-footprint measurements at the 1,000-user scale.research paper · Sep 2, 2026
Related questions
How can models adapt to low-resource languages without damaging source-language and related-task performance?How can sparse autoencoder features be shared across language models without per-model retraining?How can language models sparsely access a growing internal state without limiting it to fixed-size memory?How can video facial-expression recognition personalize vision-language models under shifts without costly test-time optimization?