Searcharxiv⌕ Search

arXiv subjects

Tun Lu

Publications and source records attributed to Tun Lu.

At least 19 recordsLinked to original sources

IntPro: Context-Aware Intent Inference via Selective Retrieval from User History

Accurately understanding user intent in context is essential for large language models (LLMs) to provide effective assistance. However, observed behavior may support multiple plausible intents, making the current context alone insufficient for identifying the user's underlying goal. A user's interaction history may contain recurring context-to-intent patterns that help disambiguate the current intent. Leveraging such historical patterns requires deciding whether to invoke retrieval for the current context and, if so, forming queries that express plausible intents rather than relying on contextual similarity alone. We present IntPro, a proxy agent that learns to either infer intent directly from the current context or generate candidate intent explanations to retrieve intent-aligned records from the user's history. We design intent explanations as a shared intent-oriented natural-language form that expresses the observed action, its functional goal, and a cue-to-intent rationale. Label-conditioned explanations represent historical records, while candidate explanations generated from the current context serve as retrieval queries in the same form. IntPro first learns direct and retrieval-conditioned inference through Direct-and-Retrieval supervised fine-tuning (SFT). IntPro then uses multi-turn GRPO with a retrieval-decision reward to jointly optimize intent prediction and selective retrieval mechanism. Experiments across three scenarios and three model backbones demonstrate that IntPro consistently improves context-aware intent inference.

cs.CL↗

LiveAsset: Learning Intent Assets from Corrective Interactions for Agent Runtime Control

Large language model (LLM) agents acting across web applications and online services often need multi-turn corrections to align their work with users' underspecified goals and preferences. In practice, users may identify problems in the current outcome without specifying the desired result, so learning user intentions requires interpreting each correction against the agent's preceding and revised outcomes. These expectations are often connected to a particular context, making it hard to know when they apply and how to ensure the agent later acts as the user expects. To this end, we propose LiveAsset, an intent harness that learns reusable intent assets from corrective interactions and uses them for runtime control of agent execution. LiveAsset jointly interprets each correction with the agent's original and revised results to infer user intent, incrementally updating the assets while preserving earlier requirements that remain applicable. The learned requirements guide agent execution from task entry and provide criteria for runtime verification, which uses the explanations to assess applicability before each reply is delivered and directs the agent to revise the result to address unmet requirements while preserving those already satisfied. Following a learning phase in which simulated users progressively reveal requirements through multi-turn corrections, we evaluate LiveAsset using same-task replay of the original requests on π-Bench, PERMA, and Vita and cross-task transfer on PAST-Bench and SkillEvolBench. Experiments show that LiveAsset better fulfills requirements learned from corrective interactions and improves performance on these benchmarks compared with memory, skill, and rule-based harness baselines. Code is available at github.com/landian60/LiveAsset.

cs.AI↗

StreamMemBench: Streaming Evaluation of Agent Memory for Future-Oriented Assistance

A central role of personal-agent memory is to turn stored information and prior interactions into future-oriented assistance. In daily use, useful cues come from what the agent observes and how the user interacts with the agent, and the agent must carry them forward from the current request to similar future tasks. Existing memory benchmarks usually test dialogue recall or task improvement in isolation, leaving the trajectory from streaming observations to later assistance largely untested. We introduce StreamMemBench, a streaming benchmark that constructs a two-step task sequence around each evidence anchor from EgoLife egocentric streams. The initial task tests evidence use, while the follow-up task tests whether feedback and interaction experience are reused. Four metrics diagnose evidence recall, initial evidence use, feedback incorporation, and follow-up reuse. Experiments with eight memory systems across two backbones show that current systems often fail to use observed evidence or turn feedback into reliable follow-up behavior, even when evidence is stored or feedback is incorporated locally. StreamMemBench is publicly available at https://github.com/landian60/StreamMemBench.

cs.AI↗

VIRA-Bench: Diagnosing Pedagogical Contingency Understanding in Online Tutoring Videos

Understanding a one-to-one tutoring video requires more than recognizing its events. A model must connect a visible teacher intervention to the learner response it elicits and to the follow-up adjusted to that response. Existing video and educational-dialogue benchmarks typically emphasize event recognition, temporal reasoning, social-state inference, or transcript-based interaction, but rarely diagnose this action-response-adjustment relation. We introduce VIRA-Bench, a focused diagnostic benchmark for short-horizon pedagogical contingency in one-to-one online tutoring. VIRA-Bench contains 1,005 tutoring clips and 1,322 multiple-choice questions with researcher evidence audit and expert verification, organized into Observable Evidence Grounding (OEG), Interaction Contingency Reasoning (ICR), and Support-Response Modeling (SRM). Across 11 models, the best reaches 61.3%, well below the 91.0% human reference on a stratified subset, and ICR generally has the lowest layer accuracy among leading models. Under controlled input manipulations, removing video most strongly degrades OEG, whereas disrupting temporal order most strongly degrades ICR. Free-response, Stem+Options, and synchronized-shuffle controls show that response format, static question cues, and video-ASR mismatch do not explain these layer-specific patterns. The results identify distinct behavioral evidence dependencies without claiming an internal architectural mechanism. VIRA-Bench provides a focused evaluation resource for multimodal dialogue grounding, pedagogical discourse, and temporal reasoning in online tutoring.

cs.HC↗

Fast-Slow Thinking RM: Efficient Integration of Scalar and Generative Reward Models

Reward models (RMs) are critical for aligning Large Language Models via Reinforcement Learning from Human Feedback (RLHF). While Generative Reward Models (GRMs) achieve superior accuracy through chain-of-thought (CoT) reasoning, they incur substantial computational costs. Conversely, Scalar Reward Models (SRMs) offer efficiency but suffer from limited performance and adaptability in complex scenarios. We introduce Fast-Slow Thinking Reward Models (F/S-RM), a hybrid RM architecture inspired by Dual Process Theory. It trains a single model to integrate two distinct reward paradigms: scalar-style first-token pairwise judgment (fast thinking) and CoT-based judgment (slow thinking), regulated by a dual-confidence activation mechanism that determines when to activate slow thinking. Under hybrid inference, F/S-RM achieves state-of-the-art accuracy with an average score of 84.3 across benchmarks, while reducing token consumption by 22.5%.

cs.CL↗

Dynamic Feature-Embedding Communication via Codebook Distillation for Federated Recommendation

Federated recommendation systems commonly protect user privacy by keeping user parameters on local devices, while exchanging item parameters for collaborative model training. However, such item parameters usually model items independently and suffer from both efficiency and effectiveness challenges, making communication costs grow with the item space and limiting cross-item generalization and robustness to noisy feedback. To address these limitations, we propose to model items via shared latent feature embeddings for communication. Residual Quantization (RQ) provides a natural way to instantiate this communication by representing each item with a short sequence of discrete code IDs, i.e., Semantic IDs (SIDs). However, directly applying centralized and static RQ-based recommendation to federated learning is non-trivial due to 1) private and biased historical interactions and 2) evolving collaborative information. We propose RQFedRec, an RQ-based federated recommendation framework for dynamic feature-embedding communication. To construct globally aligned codebooks without accessing private interactions, RQFedRec introduces an information distillation module. Each client first learns item embeddings that encode local collaborative information from private interactions, and then distills such information into feature-indexed codebooks under globally shared SIDs, making sparse and biased local signals more compatible with server aggregation. To adapt to evolving collaborative information, RQFedRec introduces a self-refining SID update module that dynamically refines global SID assignments from aggregated codebooks. Extensive experiments demonstrate that RQFedRec improves recommendation performance and reduces communication costs without relying on semantic information, while further benefiting from public semantics when available. Code is available at https://github.com/Mingzhe-Han/RQFedRec.

cs.IR↗

FedHUR: Learning Hierarchical Utility-Guided Client Relations for Personalized Federated Recommendation

Federated recommendation enables collaborative model training while keeping user interaction data on local clients. A central problem in federated recommendation is how to aggregate useful information across clients for personalized recommendation. Existing personalized aggregation methods usually construct client relations from predefined parameter-based assumptions, such as parameter similarity or complementarity, and use these relations to determine aggregation weights. However, such methods construct a single global relation, which is insufficient to capture the hierarchical and multi-granularity nature of user relations in recommendation. Moreover, these predefined relations cannot directly reflect whether the related clients can improve prediction performance after aggregation. To address these limitations, we propose FedHUR, a federated recommendation framework for learning hierarchical utility-guided client relations. FedHUR takes item-item filters as the object for relation construction and aggregation. Specifically, it first aggregates and clusters each client's local information to obtain global hierarchical information. Each client computes hierarchical utility signals based on its local information and the global hierarchical information, indicating which collaborative information is useful for improving its prediction. The server uses these utility signals to retrieve clients that are useful to that client for further personalized aggregation. Extensive experiments on five real-world datasets show that FedHUR consistently outperforms existing federated recommendation baselines, demonstrating the effectiveness of hierarchical utility-guided client relation learning. Code is available at https://github.com/Mingzhe-Han/FedHUR.

cs.IR↗

ElderBench: Benchmarking Autonomous Mobile Agents for Older Adults

While autonomous mobile agents hold great potential for assisting older adults with smartphone usage, existing GUI benchmarks mainly rely on explicit, goal-oriented instructions and rarely capture the naturally occurring language patterns of older users, such as indirect speech, referential ambiguity, and under-specified requests. This mismatch between benchmark instructions and real-world elderly interactions may hinder reliable agent deployment. To address this gap, we present ElderBench, the first benchmark for evaluating mobile GUI agents in authentic elderly-oriented scenarios. ElderBench is constructed from 249 naturally elicited smartphone tasks collected from older adults across 20 applications. We first characterize the linguistic divergence between elderly instructions and existing GUI benchmark instructions from syntactic, semantic, and pragmatic perspectives. We then evaluate mainstream GUI agents and Vision-Language Models under both online and offline settings, revealing substantial performance degradation when handling elderly-oriented instructions. Through controlled instruction normalization, failure analysis, and fine-grained linguistic feature analysis, we further identify how elderly-specific language patterns contribute to agent failures. Our findings provide actionable design insights toward more adaptive, interpretable, and age-inclusive GUI agents for older adults.

cs.AI↗

The Curse and Blessing of Mean Bias in FP4-Quantized LLM Training

FP4 training promises substantial memory and compute savings for large language models, but remains fragile because blockwise quantization is dictated by extreme activation magnitudes, which inflate dynamic range and compress long-tail signals. We identify a counterintuitive source of this failure: dominant activation outliers are not merely arbitrary sparse events, but are largely induced by a coherent rank-one mean bias, whose direction aligns with the leading anisotropic spectral component. This mean component strengthens during training, is amplified and reshaped by attention and FFN operators, and increasingly dominates top activation magnitudes. Crucially, this discovery reveals that a seemingly complex outlier-suppression problem admits a truly simple solution: isolate the coherent mean before quantization. We therefore propose Averis, a mean-residual splitting quantization method that separates the mean component using only reductions and elementwise subtractions before FP4 quantization. Across Qwen3 0.6B Dense trained on 100B tokens and Qwen3 7B A1.5B MoE trained on 50B tokens, Averis enables robust W4A4G4 FP4 training, reducing BF16 loss gaps to 1.19%/0.81% versus 2.05%/1.10% for NVIDIA's recently released Hadamard-based outlier-smoothing method, while limiting downstream gaps to 0.89/0.71 points. With only 2.20% end-to-end overhead over vanilla NVFP4, about 30% of NVIDIA's Hadamard-based design, Averis provides a hardware-efficient path to stable low-bit LLM training. Complementary to Hadamard, Averis further reduces the Qwen3-0.6B loss and downstream gaps to 0.94% and 0.73 points when combined. Code is available at: https://anonymous.4open.science/r/averis-504D.

cs.LG↗

From Hidden Profiles to Governable Personalization: Recommender Systems in the Age of LLM Agents

Personalization has traditionally depended on platform-specific user models that are optimized for prediction but remain largely inaccessible to the people they describe. As LLM-based assistants increasingly mediate search, shopping, travel, and content access, this arrangement may be giving way to a new personalization stack in which user representation is no longer confined to isolated platforms. In this paper, we argue that the key issue is not simply that large language models can enhance recommendation quality, but that they reconfigure where and how user representations are produced, exposed, and acted upon. We propose a shift from hidden platform profiling toward governable personalization, where user representations may become more inspectable, revisable, portable, and consequential across services. Building on this view, we identify five research fronts for recommender systems: transparent yet privacy-preserving user modeling, intent translation and alignment, cross-domain representation and memory design, trustworthy commercialization in assistant-mediated environments, and operational mechanisms for ownership, access, and accountability. We position these not as isolated technical challenges, but as interconnected design problems created by the emergence of LLM agents as intermediaries between users and digital platforms. We argue that the future of recommender systems will depend not only on better inference, but on building personalization systems that users can meaningfully understand, shape, and govern.

cs.IR↗

Transparent and Controllable Recommendation Filtering via Multimodal Multi-Agent Collaboration

While personalized recommender systems excel at content discovery, they frequently expose users to undesirable or discomforting information, highlighting the critical need for user-centric filtering tools. Current methods leveraging Large Language Models (LLMs) struggle with two major bottlenecks: they lack multimodal awareness to identify visually inappropriate content, and they are highly prone to "over-association" -- incorrectly generalizing a user's specific dislike (e.g., anxiety-inducing marketing) to block benign, educational materials. These unconstrained hallucinations lead to a high volume of false positives, ultimately undermining user agency. To overcome these challenges, we introduce a novel framework that integrates end-to-cloud collaboration, multimodal perception, and multi-agent orchestration. Our system employs a fact-grounded adjudication pipeline to eliminate inferential hallucinations. Furthermore, it constructs a dynamic, two-tier preference graph that allows for explicit, human-in-the-loop modifications (via Delta-adjustments), explicitly preventing the algorithm from catastrophically forgetting fine-grained user intents. Evaluated on an adversarial dataset comprising 473 highly confusing samples, the proposed architecture effectively curbed over-association, decreasing the false positive rate by 74.3% and achieving nearly twice the F1-Score of traditional text-only baselines. Additionally, a 7-day longitudinal field study with 19 participants demonstrated robust intent alignment and enhanced governance efficiency. User feedback confirmed that the framework drastically improves algorithmic transparency, rebuilds user control, and alleviates the fear of missing out (FOMO), paving the way for transparent human-AI co-governance in personalized feeds.

cs.IR↗

Drift-Aware Continual Tokenization for Generative Recommendation

Generative recommendation commonly adopts a two-stage pipeline in which a learnable tokenizer maps items to discrete token sequences (i.e. identifiers) and an autoregressive generative recommender model (GRM) performs prediction based on these identifiers. Recent tokenizers further incorporate collaborative signals so that items with similar user-behavior patterns receive similar codes, substantially improving recommendation quality. However, real-world environments evolve continuously: new items cause identifier collision and shifts, while new interactions induce collaborative drift in existing items (e.g., changing co-occurrence patterns and popularity). Fully retraining both tokenizer and GRM is often prohibitively expensive, yet naively fine-tuning the tokenizer can alter token sequences for the majority of existing items, undermining the GRM's learned token-embedding alignment. To balance plasticity and stability for collaborative tokenizers, we propose DACT, a Drift-Aware Continual Tokenization framework with two stages: (i) tokenizer fine-tuning, augmented with a jointly trained Collaborative Drift Identification Module (CDIM) that outputs item-level drift confidence and enables differentiated optimization for drifting and stationary items; and (ii) hierarchical code reassignment using a relaxed-to-strict strategy to update token sequences while limiting unnecessary changes. Experiments on three real-world datasets with two representative GRMs show that DACT consistently achieves better performance than baselines, demonstrating effective adaptation to collaborative evolution with reduced disruption to prior knowledge. Our implementation is publicly available at https://github.com/HomesAmaranta/DACT for reproducibility.

cs.IR↗

SparkTales: Facilitating Cross-Language Collaborative Storytelling through Coordinator-AI Collaboration

Cross-language collaborative storytelling plays a vital role in children's language learning and cultural development, fostering both expressive ability and intercultural awareness. Yet, in practice, children's participation is often shallow, and facilitating such sessions places heavy cognitive and organizational burdens on coordinators, who must coordinate language support, maintain children's engagement, and navigate cultural differences. To address these challenges, we conducted a formative study with coordinators to identify their needs and pain points, which guided the design of SparkTales, an intelligent support system for cross-language collaborative storytelling. SparkTales leverages both individual and common characteristics of participating children to provide coordinators with story frameworks, diverse questions, and comprehension-oriented materials, aiming to reduce coordinators' workload while enhancing children's interactive engagement. Evaluation results show that SparkTales not only significantly increases coordinators' efficiency and quality of guidance but also improves children's participation, providing valuable insights for the design of future intelligent systems supporting cross-language collaboration.

cs.HC↗

AdAEM: An Adaptively and Automated Extensible Measurement of LLMs' Value Difference

Assessing Large Language Models'(LLMs) underlying value differences enables comprehensive comparison of their misalignment, cultural adaptability, and biases. Nevertheless, current value measurement methods face the informativeness challenge: with often outdated, contaminated, or generic test questions, they can only capture the orientations on comment safety values, e.g., HHH, shared among different LLMs, leading to indistinguishable and uninformative results. To address this problem, we introduce AdAEM, a novel, self-extensible evaluation algorithm for revealing LLMs' inclinations. Distinct from static benchmarks, AdAEM automatically and adaptively generates and extends its test questions. This is achieved by probing the internal value boundaries of a diverse set of LLMs developed across cultures and time periods in an in-context optimization manner. Such a process theoretically maximizes an information-theoretic objective to extract diverse controversial topics that can provide more distinguishable and informative insights about models' value differences. In this way, AdAEM is able to co-evolve with the development of LLMs, consistently tracking their value dynamics. We use AdAEM to generate novel questions and conduct an extensive analysis, demonstrating our method's validity and effectiveness, laying the groundwork for better interdisciplinary research on LLMs' values and alignment. Codes and the generated evaluation questions are released at https://github.com/ValueCompass/AdAEM.

cs.CY↗

Echoes of Norms: Investigating Counterspeech Bots' Influence on Bystanders in Online Communities

Counterspeech offers a non-repressive approach to moderate hate speech in online communities. Research has examined how counterspeech chatbots restrain hate speakers and support targets, but their impact on bystanders remains unclear. Therefore, we developed a counterspeech strategy framework and built \textit{Civilbot} for a mixed-method within-subjects study. Bystanders generally viewed Civilbot as credible and normative, though its shallow reasoning limited persuasiveness. Its behavioural effects were subtle: when performing well, it could guide participation or act as a stand-in; when performing poorly, it could discourage bystanders or motivate them to step in. Strategy proved critical: cognitive strategies that appeal to reason, especially when paired with a positive tone, were relatively effective, while mismatch of contexts and strategies could weaken impact. Based on these findings, we offer design insights for mobilizing bystanders and shaping online discourse, highlighting when to intervene and how to do so through reasoning-driven and context-aware strategies.

cs.HC↗

Power Echoes: Investigating Moderation Biases in Online Power-Asymmetric Conflicts

Online power-asymmetric conflicts are prevalent, and most platforms rely on human moderators to conduct moderation currently. Previous studies have been continuously focusing on investigating human moderation biases in different scenarios, while moderation biases under power-asymmetric conflicts remain unexplored. Therefore, we aim to investigate the types of power-related biases human moderators exhibit in power-asymmetric conflict moderation (RQ1) and further explore the influence of AI's suggestions on these biases (RQ2). For this goal, we conducted a mixed design experiment with 50 participants by leveraging the real conflicts between consumers and merchants as a scenario. Results suggest several biases towards supporting the powerful party within these two moderation modes. AI assistance alleviates most biases of human moderation, but also amplifies a few. Based on these results, we propose several insights into future research on human moderation and human-AI collaborative moderation systems for power-asymmetric conflicts.

cs.HC↗

SD-MoE: Spectral Decomposition for Effective Expert Specialization

Mixture-of-Experts (MoE) architectures scale Large Language Models via expert specialization induced by conditional computation. In practice, however, expert specialization often fails: some experts become functionally similar, while others functioning as de facto shared experts, limiting the effective capacity and model performance. In this work, we analysis from a spectral perspective on parameter and gradient spaces, uncover that (1) experts share highly overlapping dominant spectral components in their parameters, (2) dominant gradient subspaces are strongly aligned across experts, driven by ubiquitous low-rank structure in human corpus, and (3) gating mechanisms preferentially route inputs along these dominant directions, further limiting specialization. To address this, we propose Spectral-Decoupled MoE (SD-MoE), which decomposes both parameter and gradient in the spectral space. SD-MoE improves performance across downstream tasks, enables effective expert specialization, incurring minimal additional computation, and can be seamlessly integrated into a wide range of existing MoE architectures, including Qwen and DeepSeek.

cs.LG↗

Multi-Head Attention as a Source of Catastrophic Forgetting in MoE Transformers

Mixture-of-Experts (MoE) architectures are often considered a natural fit for continual learning because sparse routing should localize updates and reduce interference, yet MoE Transformers still forget substantially even with sparse, well-balanced expert utilization. We attribute this gap to a pre-routing bottleneck: multi-head attention concatenates head-specific signals into a single post-attention router input, forcing routing to act on co-occurring feature compositions rather than separable head channels. We show that this router input simultaneously encodes multiple separately decodable semantic and structural factors with uneven head support, and that different feature compositions induce weakly aligned parameter-gradient directions; as a result, routing maps many distinct compositions to the same route. We quantify this collision effect via a route-wise effective composition number $N_{eff}$ and find that higher $N_{eff}$ is associated with larger old-task loss increases after continual training. Motivated by these findings, we propose MH-MoE, which performs head-wise routing over sub-representations to increase routing granularity and reduce composition collisions. On TRACE with Qwen3-0.6B/8B, MH-MoE effectively mitigates forgetting, reducing BWT on Qwen3-0.6B from 11.2% (LoRAMoE) to 4.5%.

cs.LG↗