CS Paper Radar

Using Context Is Not Enough: Test-Time Training for Personalized Reward Modeling

Bohao Wang, Xiaoyan Zhao, Yang Zhang, Jinghang Guo, Chun Chen, Can Wang, Jiawei Chen

cs.AI ·submitted 2026-09-28 13:13 UTC · v1

Abstract

Reinforcement learning from human feedback (RLHF) aligns large language models (LLMs) with human preferences, yet most pipelines learn a single reward model that overlooks individual differences in preferences. Personalized reward models (PRMs) address this by conditioning rewards on user-specific feedback, most commonly through in-context learning (ICL), where a user's historical comparisons are supplied as contextual preference pairs. However, we identify a key limitation of ICL-based PRMs: they fail to capture the preference relations conveyed by contextual pairs. To address this, we propose Preference-Aligned Test-Time Training (P-TTT), which explicitly encodes these relations into user-specific fast weights for personalized reward prediction. P-TTT introduces sequence-level update and apply operations to match the response-level granularity of preference feedback, together with a preference-aligned objective that directly uses pairwise preference relations to guide fast-weight adaptation. Notably, P-TTT is simple to implement and computationally efficient, updating fast weights within a single forward pass without inference-time backpropagation. Extensive experiments show that P-TTT more effectively captures historical preference relations and outperforms state-of-the-art methods by a large margin.

Closest papers

Nearest neighbours by meaning across the whole index.

Learning Preference Adaptation for Large Language Model Personalization via Verbal Reinforcement Learning

cs.CL cs.AI ·2026-08-10 ·Yuting Liu, Wei Wu, Jianzhe Zhao, Guibing Guo
similarity 0.70
Abstract

Natural language user preferences provide an interpretable interface for LLM personalization. However, universal preference summaries often contain information irrelevant to a particular downstream task. Directly supplying the full preference summary therefore wastes context capacity and introduces cross-task distraction, while manually designing task-specific preference views is difficult to scale. In this work, we study \emph{task-specific preference adaptation}: given a universal user preference summary and a downstream task, derive a task-conditioned representation that preserves sufficient decision-relevant evidence while removing redundant context. To this end, we propose \textsc{AlignXada}, a training-free meta-learning framework that induces reusable textual refinement policies for adapting universal preference summaries to task-specific ones. The refinement policy is iteratively optimized by a meta learner through verbal reinforcement learning. Across 13 tasks and three downstream models (39 task--model cells), \textsc{AlignXada} achieves an average gain of 3.82 points, improving 33 cells while retaining only 22.8\% of the original profile tokens and outperforming RAG in 36 cells. An extended faithfulness analysis further shows that the refined profiles remain largely grounded in the source preferences while preserving task-relevant personalization signals, suggesting that profile-side adaptation serves as a practical complement to universal memory construction for lifelong personalized agents.

0.04

Memory-Aware Reward Modeling: Learning Human Preferences for LLM Memory Use

cs.CL cs.AI ·2026-10-22 ·Yijiang River Dong, Zexue He, Cheng Yu, Tiancheng Hu, Michael J Ryan +7 more
similarity 0.70
Abstract

Memory enables language model assistants to maintain continuity across sessions and generate personalized responses, but knowing more about a user does not necessarily produce responses they prefer. Existing evaluations assess memory capabilities without directly establishing whether memory use benefits the user. To study this gap, we collect approximately 1,000 pairwise judgments from 87 participants using their own ChatGPT queries and personal memories. We find that memory-conditioned responses are not consistently preferred over responses generated without memory and in some cases, users strongly prefer the latter. Automatic evaluators also show limited agreement with users and favor memory-conditioned responses substantially more often than users do. Motivated by these gaps, we introduce a memory-aware reward model that learns from user judgments, achieving 72.8\% preference prediction accuracy and outperforming the existing baselines. Our findings motivate evaluation and optimization objectives that align memory use with the preferences of the people whose information is being used.

0.30

Personalize at Test Time: Learning User Preferences for Image Generation

cs.CV cs.AI ·2026-10-06 ·Jiamu Bai, Jiaming Hu, Yanhong Wu, Zellux Wang
similarity 0.68
Abstract

Diffusion models can generate high-quality images, yet aligning their outputs with individual user preferences remains challenging. A key bottleneck is accurately modeling diverse user preferences from limited feedback. Existing approaches often rely on labor-intensive manual preference annotations or vision-language models (VLM) to extract preference information from user interaction histories, introducing substantial annotation or computational costs that limit scalability. We propose an approach that learns personalized reward models directly from users' historical image preference pairs. First, we use an autoencoder to compress hundreds of visual attributes into 50 attribute-anchored preference dimensions and train an evaluator to score images along these dimensions. We then represent each user's preferences as a linear combination of the shared dimension scores, estimating the user-specific weights by maximizing the likelihood of their observed pairwise preferences under the Bradley-Terry model. This formulation reduces per-user adaptation to optimizing a low-dimensional weight vector, simplifying optimization and enabling data-efficient personalization from sparse feedback. The learned personalized rewards guide image generation at inference time while keeping the diffusion model frozen. Experiments on real-user preference data show that our approach achieves approximately 77% held-out pairwise preference prediction accuracy and improves the alignment of generated images with individual user preferences.

0.31

PerSPL: Personalized Semi-supervised Preference Learning

cs.LG ·2026-10-21 ·Xiaoan Xu, Tingzhou Wei, Juncheng Dong, Korel Gundem, Ethan X. Fang +1 more
similarity 0.68
Abstract · A coauthor has explicitly requested withdrawal due to unresolved concerns about the manuscript. In particular, Section 4 does not adequately…

Personalized reward models require user-specific preference comparisons, but collecting such feedback is costly and provides only limited supervision for each user. At the same time, large pools of responses can be generated without user identities or preference labels, yet existing personalized reward models leave this data unused. Directly learning from these responses is difficult because pairwise objectives constrain relative scores within a prompt, whereas pseudo-labeling individual responses requires scores that are comparable across prompts for the same user. We propose \textbf{PerSPL} (Personalized Semi-supervised Preference Learning), a backbone-agnostic framework that uses pair-mean score centering to stabilize this cross-prompt score coordinate and iteratively turns unlabeled responses into user-specific training signals. Against seven representative baselines across four benchmarks---PersonalLLM, HelpSteer2, Reddit TLDR, and Chatbot Arena---PerSPL achieves the highest reported mean accuracy among the evaluated methods for both seen and unseen users. The two latter benchmarks provide real-user preference comparisons in summarization and open-domain chatbot interaction, respectively. These results indicate that unlabeled responses can provide additional supervision for personalized reward modeling when per-user preference labels are scarce.

0.13

Steerable Multi-Objective Group-Relative Preference Optimization with Jointly Trained Low-Rank Adapters

cs.CL cs.AIcs.LG ·2026-11-06 ·Constantinos Karouzos, Nikolaos Aletras
similarity 0.68
Abstract · Preprint. Under review

Language model deployment often requires different trade-offs between competing objectives as user preferences change, yet standard post-training methods typically produce a single policy for a fixed objective weighting. We introduce LambdaMix, a new parameter efficient method for learning a preference conditioned family of policies with group-relative online reinforcement learning. LambdaMix jointly trains one low-rank adapter per objective over a frozen base model. One preference vector selects the adapter mixture that generates responses and weights the separately normalized rewards used to update the adapters. A single training run supports changing preferences at inference without further optimization. We experiment with four models on helpfulness and harmlessness and on solving math problems while following instructions. We also evaluate the two largest models on rewriting comments to preserve wording while reducing clues about personal attributes. On the first two tasks, LambdaMix achieves higher hypervolume than all evaluated conditioned and policy merging baselines across all four models. LambdaMix uses fewer training completions and stores fewer adapter parameters than a set of GRPO specialist policies. With architecture and initialization fixed, varying adapter mixtures and objective weights together gives wider score ranges and higher coverage than varying either alone.

0.23

COPE: Continual Personalization of LLMs under Sparse User Feedback via User Embeddings and Self-Evaluation

cs.LG cs.AI ·2026-09-22 ·Ruike Cao, Fugen Yao, Liang Dong, Jian Xu, Guanjun Jiang +1 more
top venue similarity 0.68
Abstract

While Large Language Models (LLMs) have achieved remarkable results across various benchmarks, their alignment with normative values often results in homogenized responses that fail to address diverse user preferences. Existing training-free methods often occupy valuable context windows through prompt engineering, while training-based methods typically remain static post-training, failing to support the continual optimization required in real-world settings. To address these challenges, we propose COPE (Continual Optimization with Personalized embedding and self-Evaluation), a novel optimization framework tailored for real-world-motivated interaction settings with sparse user feedback. Our framework assigns learnable personalized embeddings to each user and synergistically integrates preference capture, self-evaluation calibration, and personalized response optimization within a single update step. A key innovation of our method is the use of self-evaluation to generate proxy rewards, enabling continuous model updates even when explicit user feedback is unavailable. Experiments show that COPE consistently outperforms strong training-free and training-based baselines under sparse feedback, and remains complementary to Retrieval-Augmented Prompting (RAP). Further analyses confirm COPE's reliable self-evaluation, meaningful preference patterns, stable general capabilities, and robustness under shifting preferences and alternative evaluators.

0.34

Judging in Latent Space: Efficient Generative Reward Modeling via Semantics-Preserving Compression

cs.CL ·2026-10-07 ·Mingqing Yuan, Xiaobo Liang, Junwei Yang, Ziwei Chen, Zeren Zhang +3 more
benchmark similarity 0.68
Abstract

Reward modeling often requires jointly representing and reasoning over multiple evaluation criteria, yet verbalizing this process token by token can incur substantial inference cost. Recent work on latent reasoning suggests that continuous states may support this computation more compactly. We introduce LatentGRM, a latent evaluation framework built on semantic chunking, compression, and reconstruction. By using the structure of rubric-guided evaluations to guide compression, LatentGRM learns compact continuous trajectories that support autonomous pairwise judgments without generating textual assessments. A separate interpreter reconstructs evaluation text from these trajectories, providing an offline view of the information retained under compression. Under matched training data and backbones, LatentGRM achieves competitive aggregate preference accuracy relative to explicit Supervised Fine-Tuning (SFT) judges at both 4B and 8B scales. Across four benchmark domains, LatentGRM-8B compresses evaluation trajectories by 8.9--9.2x and reduces total judge inference time by 6.1--7.0x at vote@5. Controlled rubric interventions show that criterion-dependent preference information is carried through the latent sequence. Together, these results demonstrate that continuous latent evaluation can substantially reduce inference cost while preserving competitive judgment quality.

0.15

Decision-Focused Reinforcement Learning for LLM Post-Training

cs.LG cs.AI ·2026-11-19 ·Haotian Zhai, Leqi Liu
similarity 0.66
Abstract

Large language models (LLMs) have shown promise as predictors across various tasks, benefiting from their ability to use information from unstructured context. However, when predictions serve as parameters of downstream optimization problems, higher prediction accuracy does not necessarily translate into lower decision costs. We propose decision-focused reinforcement learning (DFRL), a framework for post-training LLMs to predict optimization parameters using negative downstream decision regret as reward. DFRL trains LLM predictors to minimize the costs of the decisions induced by their predictions, without requiring downstream convexity or differentiability, and integrates readily into existing RL post-training frameworks. To address expensive reward evaluation, we develop two approaches. Heuristic warm-starting uses inexpensive approximate rewards during an initial phase before switching to refinement with exact rewards. Reward correction combines inexpensive estimates with randomized exact evaluations, preserving the expected policy gradient while reducing the expected number of exact solves per update. Experiments on three synthetic optimization problems demonstrate improvements over prediction-loss baselines, while a larger knapsack experiment shows faster early improvement under short training budgets. On inventory planning with H&M transactions and product descriptions, our best configuration, in which the LLM reasons before predicting, reduces regret by 31.4% relative to the SFT baseline and 12.4% relative to the strongest evaluated conventional forecasting baseline.

0.17