Tag
The paper introduces MIRROR, a self-distillation framework from Baidu that shifts LLM personalization from reference-token imitation to internalizing user preferences via on-policy distributional alignment, with a MIRROR-F focal plug-in that improves content quality while reducing catastrophic forgetting across benchmarks and model scales.
The paper proposes three approaches for few-shot author-style adaptation of LLMs—contrastive activation steering, a steering-vector prediction network, and a hypernetwork that predicts LoRA adapters—and finds the hypernetwork offers the best trade-off between style imitation and output quality for both seen and unseen authors.
The article discusses personalization features across various LLMs like Claude, Gemini, and Copilot, providing a guide on custom instructions to improve response accuracy and reduce hallucinations.
HyperTrace is a training-free framework for online LLM personalization that uses interpretable natural-language hypotheses to trace latent user preferences, improving response alignment and consistency.
HypReflect is a framework for continual personalization in LLM assistants that infers explicit preference hypotheses from user signals and uses self-distillation to adapt, outperforming baselines and generalizing to unseen users and cross-domain settings.
PersonalBench is a new benchmark that evaluates inference-time personalization methods in LLMs through authorship verification, LLM-as-judge, and stylometrics, finding that while methods produce author-differentiated output, they do not bridge the gap to human authorship.
This paper introduces PAC-Bayes-regularized Meta-LoRA for cross-domain LLM personalization, enabling zero- and few-shot adaptation to user preferences while preventing overfitting under sparse evidence. Experiments on benchmarks like HiCUPID show significant improvements in cross-domain win rates and cold-start scenarios.
This paper introduces MirageBench, a benchmark showing that LLMs with persistent memory fabricate user profiles through over-inference 35-49% of the time, and reveals that model self-reported confidence is inversely correlated with actual over-inference, making self-monitoring unreliable for comparing models.
This paper introduces a decoupled evaluation paradigm to separate memory recall from actual utilization in LLM personalization, finding that agents often recall user preferences but fail to act on them in behavioral scenarios. Large-scale experiments across 16 systems and five memory architectures reveal a significant Know-Act gap, especially for health-related preferences.
This essay proposes 'Guardian Angels,' highly personalized LLMs that emulate a user's values and preferences to boost productivity and provide security against advanced AI threats, using techniques like dynamic evaluation and active learning.
This paper studies the gap between synthetic and human data for evaluating LLM personalization across three stages: attribute extraction, relevance matching, and response generation. Results show models perform worse on real human data, and the authors introduce lightweight training interventions to improve alignment.
This paper investigates the effectiveness of LLM personalization by putting real humans back into the evaluation loop, revealing systematic gaps between human judgments and LLM outputs at every stage of the personalization pipeline, and highlighting the limitations of synthetic data and LLM judges.
This paper proposes PUMA, a framework for LLM personalization in multi-turn conversations that models latent user states and uses the Free Energy Principle to select dialogue actions, improving long-horizon outcomes on healthcare counseling benchmarks.
This paper introduces Test-Time Personalization (TTP), a framework that improves LLM personalization by scaling inference-time computation through candidate sampling and reward-based selection. It diagnoses failure modes in standard reward models and proposes a probabilistic personalized reward model to mitigate them.
This paper introduces IRiS, a training-free framework for situational personality steering in LLMs that moves beyond static persona modeling by identifying and leveraging situation-dependent persona neurons. The approach demonstrates that LLM behavior varies contextually and proposes neuron-based identification, retrieval, and weighted steering methods validated on PersonalityBench and a new SPBench benchmark.