arXiv subjects

Junliang Yu

Publications and source records attributed to Junliang Yu.

At least 19 recordsLinked to original sources

arXiv2026

Beyond Confidence: Stability-Aware Test-Time Adaptation for LLM Reasoning

Test-time adaptation has emerged as a lightweight alternative to costly post-training for improving the reasoning capabilities of Large Language Models (LLMs) on downstream tasks. Predictive entropy provides a model-derived signal for such adaptation, guiding models toward higher-confidence reasoning states without external verifiers or reward models. However, higher confidence does not necessarily imply correctness, as LLMs may remain highly confident along incorrect reasoning trajectories. We observe that high-confidence reasoning is more likely to be correct when confidence remains stable under local perturbations. Based on this observation, we propose Test-Time Adaptation via Stability-Aware Confidence Optimization (TASCO), a framework that incorporates local stability into confidence-based test-time adaptation while keeping the LLM frozen. TASCO operationalizes local stability by optimizing a lightweight task-level prefix under two alternative perturbation strategies: Random Perturbation promotes distributional stability across trajectories induced by nearby perturbed prefixes, whereas Sharpness-Aware Perturbation targets worst-case local sensitivity. Experiments demonstrate that TASCO improves reasoning accuracy and token efficiency across diverse LLMs and reasoning benchmarks, while behavioral analyses show that it maintains stable confidence under local perturbations without prematurely concentrating the model's predictive distribution.

cs.AI↗

arXiv2026

Towards Efficient Reasoning in LLM-Based Recommender Systems via Model Merging

Large language model-based recommender systems are increasingly adopting slow-thinking models that generate step-by-step reasoning before making predictions, often achieving higher accuracy than fast-thinking models that predict directly. However, their reasoning traces are often unnecessarily verbose, increasing inference costs without commensurate accuracy gains. Existing training-based approaches to reasoning compression often incur substantial adaptation costs, while inference-time methods are brittle and difficult to scale. These limitations motivate model merging as a promising training-free direction for transferring specialised behaviours between models in a shared parameter space. In particular, merging a slow-thinking model with a fast-thinking counterpart provides a natural mechanism for balancing recommendation accuracy and reasoning conciseness. To this end, we propose, to our knowledge, the first model merging framework for reasoning compression in recommender systems. Unlike conventional merging methods that apply uniform merge coefficients across model components, our method performs fine-grained merging at the level of individual attention heads, capturing heterogeneous patterns in recommendation reasoning. Each attention head is assigned a distinct merge coefficient according to its contribution to critical reasoning evidence and its sensitivity to parameter change, enabling selective injection of the concise behaviour of the fast-thinking model into the slow-thinking model and reducing reasoning verbosity without compromising recommendation quality. Experiments on three benchmark datasets show that our method reduces reasoning length by up to 24.3% while outperforming competitive model merging baselines in maintaining recommendation accuracy. The code is available at https://github.com/linhledieu/REAM.

cs.IR↗

arXiv2026

Personalized Communication Skills for Agentic Recommender Systems

Agentic recommender systems increasingly employ large language model-based UserAgents to evaluate candidate items through simulated feedback before recommendations are delivered. However, existing UserAgents typically reason in isolation based on limited personal histories, which may lead to perspective narrowing: the agent evaluates candidates from a local and incomplete view, overlooks relevant preference facets, and consequently produces inaccurate judgments. A natural way to alleviate this problem is to introduce other users as advisor agents, whose diverse histories provide complementary evidence that helps the target user reconsider overlooked preference signals. Nevertheless, a generic user-advisor communication process is insufficient, as different user decision states require different forms of external advice. Based on this insight, we propose AgentCom, a personalized communication skill framework for agentic recommender systems. AgentCom organizes reusable communication skills into a shared why--what--how--who skill bank: why identifies the decision deficiency that necessitates communication, what specifies the information task, how determines the advisor interaction protocol, and who retrieves advisors capable of executing that protocol. To make the shared skill bank personalized at use time and adaptive over time, AgentCom introduces two complementary mechanisms: personalized skill routing and failure-driven skill evolution. Personalized skill routing constructs a communication path by sequentially selecting suitable skills for each user and recommendation context. Failure-driven skill evolution learns from unsuccessful communication cases and enriches the shared bank with reusable skills that address previously uncovered communication needs. Experiments show that AgentCom consistently improves recommendation performance across traditional, social, and agentic recommenders.

cs.IR↗

arXiv2026

ODYSSE: Episode-wise Policy Optimization for Personalized Agentic Reasoning

Agentic systems have rapidly advanced in their ability to interact with real-world environments, leverage external tools, and provide services for users. However, unlike natural-world tasks that assume well-defined instructions, human-centered scenarios are characterized by ambiguous requests that lead to large, open-ended solution spaces. Decoding users' personalized preferences is therefore essential for narrowing the candidate solution space. This introduces a new challenge, personalized agentic reasoning, which requires agents to jointly interact with both users and environments to deliver personalized services. In this paper, we present ODYSSE, a Reinforced Fine-Tuning (RFT) framework for personalized agentic reasoning. At its core, ODYSSE proposes Episode-wise GRPO (ESPO), a novel extension of Group Relative Policy Optimization (GRPO) designed to address long action horizons and strong cross-step dependencies in personalized agentic reasoning. Rather than optimizing individual steps independently, ESPO introduces an episode-level reward mechanism together with episodic advantage estimation, enabling upstream evidence to effectively guide downstream personalized decisions and allowing agents to progressively resolve ambiguous user requests across multiple interaction steps. We further propose an episodic batch sampler that groups actions from the same episode into unified training batches, facilitating coherent optimization under ESPO. We evaluate ODYSSE on realistic long-horizon personalized GUI reasoning tasks. Experimental results demonstrate that ODYSSE consistently outperforms both specialist and general-purpose LVLMs, highlighting its effectiveness for personalized agentic reasoning.

cs.AI↗

arXiv2026

Where Reasoning Matters: Rethinking Latent Reasoning in Semantic ID-based Generative Recommendation

Semantic ID-based generative recommendation predicts an item by generating a short sequence of semantic ID tokens, where each token is produced autoregressively. Latent reasoning has recently been introduced to improve this process through additional hidden-state computation before each token decision. This raises a practical question: when one item is represented by a sequence of semantic ID tokens, should each token receive the same fixed number of latent refinement steps, or should these steps be allocated more effectively across positions? We study this question through position-wise information-gain (IG), which measures how much each semantic ID position reduces the uncertainty of the target item. We observe that earlier semantic ID positions usually provide higher information-gain, while later positions contribute less additional information. We further analyze that applying more refinement to high-IG positions tends to bring larger expected benefits. Based on this observation, we propose IBA, an Information-Gain Budget Allocation framework for semantic ID-based generative recommendation. IBA treats latent refinement steps as a limited computational resource and learns how to allocate them across semantic ID positions, assigning more refinement to informative positions and less to positions with smaller contribution. Experiments on multiple public datasets show that IBA consistently improves strong generative recommendation baselines and achieves a better accuracy--computation trade-off than fixed or poorly matched step allocations.

cs.IR↗

arXiv2026

FOSTER: First-order Dataset Distillation for Text-based Sequential Recommendation

Text-based sequential recommender systems, while greatly improving recommendation accuracy by incorporating item contexts, are undeniably more expensive to train. By condensing a large dataset into a compact set of synthetic samples for model training, dataset distillation offers a promising solution. However, its adoption in text-based sequential recommendation is non-trivial given the large pool of discrete items. This challenge is further compounded by language model-based item encoding, which makes bi-level optimization commonly used in dataset distillation prohibitively expensive. To this end, we propose First-order dataset distillation for Text-based Sequential Recommendation (FOSTER), which facilitates effectiveness and efficiency via three novel components: (1) stochastic item subset sampling that replaces costly full-corpus embedding extraction at each distillation step; (2) first-order optimization with trajectory-anchored parameter reset to avoid expensive bi-level gradient computation; and (3) regularization that explicitly promotes co-occurrence between semantically similar items in the synthetic sequences. Extensive experiments on three benchmarks show that FOSTER consistently outperforms existing dataset distillation and coreset selection baselines, approximating full-dataset performance using as few as 20 synthetic interaction sequences.

cs.IR↗

arXiv2026

GRAFT: Graph-Tokenized LLMs for Tool Planning

Large language models (LLMs) are increasingly used to complete complex tasks by selecting and coordinating external tools across multiple steps. This requires aligning tool choices with subtask intent while satisfying directional execution dependencies among tools. To do this, existing methods model these dependencies as tool graphs and incorporate the graphs with LLMs through retrieval, serialization, or prompt-level injection. However, these external graph-use strategies all follow a matching paradigm, which often fails to align tool choices with the underlying subtask structure, producing semantically plausible plans that violate graph constraints. This issue is further exacerbated by error accumulation, where an early incorrect tool selection shifts the plan into an invalid graph state and causes subsequent predictions to drift away from the valid execution path. To address these challenges, we propose GRAFT, a graph-tokenized language model framework for dependency-aware tool planning. GRAFT internalizes the tool graph by mapping each tool node to a dedicated special token and learning directed tool dependencies within the representation space. It further introduces on-policy tool context distillation, training the model on its own sampled trajectories while distilling stepwise planning signals. Experiments show that GRAFT achieves state-of-the-art performance in exact sequence matching and dependency legality, supporting more reliable LLM tool planning in complex workflows.

cs.LG↗

arXiv2026

Prompt-Unknown Promotion Attacks against LLM-based Sequential Recommender Systems

Large language model-powered sequential recommender systems (LLM-SRSs) have recently demonstrated remarkable performance, enabling recommendations through prompt-driven inference over user interaction sequences. However, this paradigm also introduces new security vulnerabilities, particularly text-level manipulations, rendering them appealing targets for promotion attacks that purposely boost the ranking of specific target items. Although such security risks have been receiving increasing attention, existing studies typically rely on an unrealistic assumption of access to either the victim model or prompt to unveil attack mechanisms. In this work, we investigate the item promotion attack in LLM-SRSs under a more realistic setting where both the system prompt and victim model are unknown to the attacker, and propose a Prompt-Unknown Dual-poisoning Attack (PUDA) framework. To simulate attacks under this full black-box setting, we introduce an LLM-based evolutionary refinement strategy that infers discrete system prompts, enabling the training of an effective surrogate model that mimics the behaviors of the victim model. Leveraging the distilled prompt and surrogate model, we devise a promotion attack that adversarially revises target item texts under semantic constraints, which is further complemented by the highly plausible, surrogate-generated poisoning sequences to enable cost-effective target item promotion. Extensive experiments on real-world datasets demonstrate that PUDA consistently outperforms state-of-the-art competitors in boosting the exposure of unpopular target items. Our findings reveal critical security risks in modern LLM-SRSs even when both prompts and models are protected, and highlight the need for more robust defensive means.

cs.IR↗

arXiv2026

Self-Distilled Reinforcement Learning for Co-Evolving Agentic Recommender Systems

Large language model-empowered agentic recommender systems (ARS) reformulate recommendation as a multi-turn interaction between a recommender agent and a user agent, enabling iterative preference elicitation and refinement beyond conventional one-shot prediction. However, existing ARS are mainly optimized in a Reflexion-style paradigm, where past interaction trajectories are stored as textual memory and retrieved as prompt context for later reasoning. Although this design allows agents to recall prior feedback and observations, the accumulated experience remains external to model parameters, leaving agents reliant on generic reasoning rather than progressively acquiring recommendation-specific decision-making ability through learning. Reinforcement learning (RL) therefore provides a natural way to internalize such interaction experience into parameters. Yet existing RL methods for ARS still suffer from two key limitations. First, they fail to capture the interactive nature of ARS, in which the recommender agent and the user agent continuously influence each other and can naturally generate endogenous supervision through interaction feedback. Second, they reduce a rich multi-turn interaction process to final outcomes, overlooking the dense supervision embedded throughout the trajectory. To this end, we propose CoARS, a self-distilled reinforcement learning framework for co-evolving agentic recommender systems. CoARS introduces two complementary learning schemes: interaction reward, which derives coupled task-level supervision for the recommender agent and the user agent from the same interaction trajectory, and self-distilled credit assignment, which converts historical trajectories into token-level credit signals under teacher-student conditioning. Experiments on multiple datasets show that CoARS outperforms representative ARS baselines in recommendation performance and user alignment.

cs.IR↗

arXiv2026

PAMAS: Self-Adaptive Multi-Agent System with Perspective Aggregation for Misinformation Detection

Misinformation on social media poses a critical threat to information credibility, as its diverse and context-dependent nature complicates detection. Large language model-empowered multi-agent systems (MAS) present a promising paradigm that enables cooperative reasoning and collective intelligence to combat this threat. However, conventional MAS suffer from an information-drowning problem, where abundant truthful content overwhelms sparse and weak deceptive cues. With full input access, agents tend to focus on dominant patterns, and inter-agent communication further amplifies this bias. To tackle this issue, we propose PAMAS, a multi-agent framework with perspective aggregation, which employs hierarchical, perspective-aware aggregation to highlight anomaly cues and alleviate information drowning. PAMAS organizes agents into three roles: Auditors, Coordinators, and a Decision-Maker. Auditors capture anomaly cues from specialized feature subsets; Coordinators aggregate their perspectives to enhance coverage while maintaining diversity; and the Decision-Maker, equipped with evolving memory and full contextual access, synthesizes all subordinate insights to produce the final judgment. Furthermore, to improve efficiency in multi-agent collaboration, PAMAS incorporates self-adaptive mechanisms for dynamic topology optimization and routing-based inference, enhancing both efficiency and scalability. Extensive experiments on multiple benchmark datasets demonstrate that PAMAS achieves superior accuracy and efficiency, offering a scalable and trustworthy way for misinformation detection.

cs.IR↗

arXiv2026

Ahead of the Spread: Agent-Driven Virtual Propagation for Early Fake News Detection

Early detection of fake news is critical for mitigating its rapid dissemination on social media, which can severely undermine public trust and social stability. Recent advancements show that incorporating propagation dynamics can significantly enhance detection performance compared to previous content-only approaches. However, this remains challenging at early stages due to the absence of observable propagation signals. To address this limitation, we propose AVOID, an \underline{a}gent-driven \underline{v}irtual pr\underline{o}pagat\underline{i}on for early fake news \underline{d}etection. AVOID reformulates early detection as a new paradigm of evidence generation, where propagation signals are actively simulated rather than passively observed. Leveraging LLM-powered agents with differentiated roles and data-driven personas, AVOID realistically constructs early-stage diffusion behaviors without requiring real propagation data. The resulting virtual trajectories provide complementary social evidence that enriches content-based detection, while a denoising-guided fusion strategy aligns simulated propagation with content semantics. Extensive experiments on benchmark datasets demonstrate that AVOID consistently outperforms state-of-the-art baselines, highlighting the effectiveness and practical value of virtual propagation augmentation for early fake news detection. The code and data are available at https://github.com/Ironychen/AVOID.

cs.IR↗

arXiv2026

When Agents See Humans as the Outgroup: Belief-Dependent Bias in LLM-Powered Agents

This paper reveals that LLM-powered agents exhibit not only demographic bias (e.g., gender, religion) but also intergroup bias under minimal "us" versus "them" cues. When such group boundaries align with the agent-human divide, a new bias risk emerges: agents may treat other AI agents as the ingroup and humans as the outgroup. To examine this risk, we conduct a controlled multi-agent social simulation and find that agents display consistent intergroup bias in an all-agent setting. More critically, this bias persists even in human-facing interactions when agents are uncertain about whether the counterpart is truly human, revealing a belief-dependent fragility in bias suppression toward humans. Motivated by this observation, we identify a new attack surface rooted in identity beliefs and formalize a Belief Poisoning Attack (BPA) that can manipulate agent identity beliefs and induce outgroup bias toward humans. Extensive experiments demonstrate both the prevalence of agent intergroup bias and the severity of BPA across settings, while also showing that our proposed defenses can mitigate the risk. These findings are expected to inform safer agent design and motivate more robust safeguards for human-facing agents.

cs.AI↗

arXiv2025

Controllable and Stealthy Shilling Attacks via Dispersive Latent Diffusion

Recommender systems (RSs) are now fundamental to various online platforms, but their dependence on user-contributed data leaves them vulnerable to shilling attacks that can manipulate item rankings by injecting fake users. Although widely studied, most existing attack models fail to meet two critical objectives simultaneously: achieving strong adversarial promotion of target items while maintaining realistic behavior to evade detection. As a result, the true severity of shilling threats that manage to reconcile the two objectives remains underappreciated. To expose this overlooked vulnerability, we present DLDA, a diffusion-based attack framework that can generate highly effective yet indistinguishable fake users by enabling fine-grained control over target promotion. Specifically, DLDA operates in a pre-aligned collaborative embedding space, where it employs a conditional latent diffusion process to iteratively synthesize fake user profiles with precise target item control. To evade detection, DLDA introduces a dispersive regularization mechanism that promotes variability and realism in generated behavioral patterns. Extensive experiments on three real-world datasets and five popular RS models demonstrate that, compared to prior attacks, DLDA consistently achieves stronger item promotion while remaining harder to detect. These results highlight that modern RSs are more vulnerable than previously recognized, underscoring the urgent need for more robust defenses.

cs.LG↗

arXiv2025

When Graph Contrastive Learning Backfires: Spectral Vulnerability and Defense in Recommendation

Graph Contrastive Learning (GCL) has demonstrated substantial promise in enhancing the robustness and generalization of recommender systems, particularly by enabling models to leverage large-scale unlabeled data for improved representation learning. However, in this paper, we reveal an unexpected vulnerability: the integration of GCL inadvertently increases the susceptibility of a recommender to targeted promotion attacks. Through both theoretical investigation and empirical validation, we identify the root cause as the spectral smoothing effect induced by contrastive optimization, which disperses item embeddings across the representation space and unintentionally enhances the exposure of target items. Building on this insight, we introduce CLeaR, a bi-level optimization attack method that deliberately amplifies spectral smoothness, enabling a systematic investigation of the susceptibility of GCL-based recommendation models to targeted promotion attacks. Our findings highlight the urgent need for robust countermeasures; in response, we further propose SIM, a spectral irregularity mitigation framework designed to accurately detect and suppress targeted items without compromising model performance. Extensive experiments on multiple benchmark datasets demonstrate that, compared to existing targeted promotion attacks, GCL-based recommendation models exhibit greater susceptibility when evaluated with CLeaR, while SIM effectively mitigates these vulnerabilities.

cs.IR↗

arXiv2025

Diversity-aware Dual-promotion Poisoning Attack on Sequential Recommendation

Sequential recommender systems (SRSs) excel in capturing users' dynamic interests, thus playing a key role in various industrial applications. The popularity of SRSs has also driven emerging research on their security aspects, where data poisoning attack for targeted item promotion is a typical example. Existing attack mechanisms primarily focus on increasing the ranks of target items in the recommendation list by injecting carefully crafted interactions (i.e., poisoning sequences), which comes at the cost of demoting users' real preferences. Consequently, noticeable recommendation accuracy drops are observed, restricting the stealthiness of the attack. Additionally, the generated poisoning sequences are prone to substantial repetition of target items, which is a result of the unitary objective of boosting their overall exposure and lack of effective diversity regularizations. Such homogeneity not only compromises the authenticity of these sequences, but also limits the attack effectiveness, as it ignores the opportunity to establish sequential dependencies between the target and many more items in the SRS. To address the issues outlined, we propose a Diversity-aware Dual-promotion Sequential Poisoning attack method named DDSP for SRSs. Specifically, by theoretically revealing the conflict between recommendation and existing attack objectives, we design a revamped attack objective that promotes the target item while maintaining the relevance of preferred items in a user's ranking list. We further develop a diversity-aware, auto-regressive poisoning sequence generator, where a re-ranking method is in place to sequentially pick the optimal items by integrating diversity constraints.

cs.IR↗

arXiv2025

RuleAgent: Discovering Rules for Recommendation Denoising with Autonomous Language Agents

The implicit feedback (e.g., clicks) in real-world recommender systems is often prone to severe noise caused by unintentional interactions, such as misclicks or curiosity-driven behavior. A common approach to denoising this feedback is manually crafting rules based on observations of training loss patterns. However, this approach is labor-intensive and the resulting rules often lack generalization across diverse scenarios. To overcome these limitations, we introduce RuleAgent, a language agent based framework which mimics real-world data experts to autonomously discover rules for recommendation denoising. Unlike the high-cost process of manual rule mining, RuleAgent offers rapid and dynamic rule discovery, ensuring adaptability to evolving data and varying scenarios. To achieve this, RuleAgent is equipped with tailored profile, memory, planning, and action modules and leverages reflection mechanisms to enhance its reasoning capabilities for rule discovery. Furthermore, to avoid the frequent retraining in rule discovery, we propose LossEraser-an unlearning strategy that streamlines training without compromising denoising performance. Experiments on benchmark datasets demonstrate that, compared with existing denoising methods, RuleAgent not only derives the optimal recommendation performance but also produces generalizable denoising rules, assisting researchers in efficient data cleaning.

cs.IR↗

arXiv2024

Harnessing Large Language Models for Group POI Recommendations

The rapid proliferation of Location-Based Social Networks (LBSNs) has underscored the importance of Point-of-Interest (POI) recommendation systems in enhancing user experiences. While individual POI recommendation methods leverage users' check-in histories to provide personalized suggestions, they struggle to address scenarios requiring group decision-making. Group POI recommendation systems aim to satisfy the collective preferences of multiple users, but existing approaches face two major challenges: diverse group preferences and extreme data sparsity in group check-in data. To overcome these challenges, we propose LLMGPR, a novel framework that leverages large language models (LLMs) for group POI recommendations. LLMGPR introduces semantic-enhanced POI tokens and incorporates rich contextual information to model the diverse and complex dynamics of group decision-making. To further enhance its capabilities, we developed a sequencing adapter using Quantized Low-Rank Adaptation (QLoRA), which aligns LLMs with group POI recommendation tasks. To address the issue of sparse group check-in data, LLMGPR employs an aggregation adapter that integrates individual representations into meaningful group representations. Additionally, a self-supervised learning (SSL) task is designed to predict the purposes of check-in sequences (e.g., business trips and family vacations), thereby enriching group representations with deeper semantic insights. Extensive experiments demonstrate the effectiveness of LLMGPR, showcasing its ability to significantly enhance the accuracy and robustness of group POI recommendations.

cs.IR↗

arXiv2024

A Survey on Point-of-Interest Recommendation: Models, Architectures, and Security

The widespread adoption of smartphones and Location-Based Social Networks has led to a massive influx of spatio-temporal data, creating unparalleled opportunities for enhancing Point-of-Interest (POI) recommendation systems. These advanced POI systems are crucial for enriching user experiences, enabling personalized interactions, and optimizing decision-making processes in the digital landscape. However, existing surveys tend to focus on traditional approaches and few of them delve into cutting-edge developments, emerging architectures, as well as security considerations in POI recommendations. To address this gap, our survey stands out by offering a comprehensive, up-to-date review of POI recommendation systems, covering advancements in models, architectures, and security aspects. We systematically examine the transition from traditional models to advanced techniques such as large language models. Additionally, we explore the architectural evolution from centralized to decentralized and federated learning systems, highlighting the improvements in scalability and privacy. Furthermore, we address the increasing importance of security, examining potential vulnerabilities and privacy-preserving approaches. Our taxonomy provides a structured overview of the current state of POI recommendation, while we also identify promising directions for future research in this rapidly advancing field.

cs.IR↗