arXiv subjects

Zhicong Lu

Publications and source records attributed to Zhicong Lu.

At least 19 recordsLinked to original sources

arXiv2026

You Really Didn't Get That? Benchmarking Social Pragmatic Inference for Indirect and Playful Chinese Online Comments

Chinese online comments often convey social meaning through indirect and playful language that is hard to interpret without context. Existing evaluations largely organize items around predefined phenomena or controlled pragmatic categories, leaving open whether models can distinguish plausible readings of what a naturally occurring comment is doing in a particular exchange. We introduce a benchmark for evaluating whether LLMs can recover such situated pragmatic meanings. From more than 200,000 public Chinese social media interaction records, we construct 4,735 human-validated diagnostic items, each pairing a target comment with reconstructed preceding context and plausible misreadings. We evaluate eight LLMs as both question writers and solvers in a cross-writer setting. The task is challenging: the strongest model achieves 81.42% leave-writer-out accuracy. Across all eight models, the mean leave-writer-out accuracy is 68.70% while human accuracy was 90.8%. Case analysis shows that models often recognize broad irony or playfulness while misidentifying the mechanism or interactional move.

cs.CL↗

arXiv2026

Balancing Safety and Autonomy: Accessibility-Oriented Interventions in Generative AI for Cognitive Impairment

Generative AI systems are increasingly used by older adults with cognitive impairment for everyday tasks such as information seeking, health management, and communication. While these systems provide flexible, language-based support, their open-ended outputs introduce risks of over-reliance, misinterpretation, and inappropriate decision-making. Prior work has focused on usability and adoption, with limited attention to how system design shapes users' participation in decision-making and the distribution of agency in care contexts. We present a qualitative study of 45 individuals with cognitive impairment and their caregivers. We identify five accessibility-oriented mechanisms: AI Capability Constraint, Human Oversight Embedding, Cognitive Engagement Maintenance, Human-AI Relationship Regulation, and Risk Transparency and Control, through which systems structure interaction. These mechanisms both support and constrain users by redistributing decision-making across users and caregivers. We show that their effects vary by impairment level: while protective mechanisms support users with severe impairment, they can restrict autonomy for those with mild impairment. As impairment progresses, tensions become less visible as user participation diminishes. Our findings highlight the need for dynamic designs that balance safety and autonomy in AI-supported care.

cs.HC↗

arXiv2026

CAVE: Competence-Aware Visual Boundary Evidence Alignment for Video Temporal Grounding

Large vision-language models (LVLMs) have achieved substantial performance gains in Video Temporal Grounding (VTG) through reinforcement learning (RL). However, existing methods primarily rely on outcome correctness rewards that evaluate only the final predicted intervals, leaving boundary-related visual evidence and its correspondence with timestamp predictions insufficiently constrained. In this paper, we delve into timestamp prediction and its underlying boundary-level visual evidence, showing prevalent misalignment between visual evidence and predicted timestamps across widely used benchmarks. To address this issue, we propose Competence-Aware Visual Boundary Evidence Alignment (CAVE), which augments localization optimization with boundary-specific visual evidence rewards to mitigate evidence-timestamp misalignment. Specifically, to explicitly represent the boundary-specific visual evidence, CAVE introduces boundary-specific evidence tokens and initializes their structured generation and distinct boundary semantics through a lightweight supervised warm-up. During RL, the visual boundary evidence alignment reward reinforces the visual attention of special evidence tokens within the ground-truth boundaries, thereby promoting alignment between visual evidence and temporal boundaries. Moreover, performance-aware gating for evidence supervision is designed to adaptively retain evidence guidance for poorly localized groups while reducing it once localization becomes sufficiently accurate to avoid over-constraining fine-grained boundary refinement. Extensive experiments on several public VTG benchmarks demonstrate the effectiveness of our method.

cs.CL↗

arXiv2026

Orthogonal Representation Editing: Decoupling Semantic Entanglement in Batch Knowledge Editing of LLMs

Knowledge editing aims to efficiently update factual information in Large Language Models (LLMs) without full retraining. However, existing methods still suffer from performance degradation in batch knowledge editing. We identify that semantic representation entanglement, such as overlapping concepts and shared syntactic patterns, accumulates interference in the representation space and reduces editing precision. To bridge this gap, in this paper, we propose Orthogonal Representation Editing (ORE), which performs edits in the hidden representation space of LLMs by constructing a general semantic subspace and enforcing orthogonal constraints on edit vectors, effectively decoupling semantic entanglement. Furthermore, we introduce a gated non-linear representation head to enable adaptive learning of editing locations and precise control over knowledge injection. Extensive experiments show that ORE outperforms existing methods and achieves superior performance in cross-lingual knowledge editing scenarios. We release our code at https://github.com/YVVH/ORE.

cs.CL↗

arXiv2026

CORA: Analyzing and bridging thinking-answer gap in Multimodal RLVR via Consistency-Oriented Reasoning Alignment

Reinforcement learning with verifiable rewards (RLVR) has successfully elicited the reasoning capabilities of large language models, motivating its extension to multimodal scenarios. Existing methods primarily focus on improving the visual coverage of reasoning traces and mitigating visual hallucinations, but underestimate the semantic inconsistency between the reasoning process and the final answer. In this paper, we delve into thinking-answer inconsistency in RLVR for large vision-language models (LVLMs), showing thorough analyses of rollouts collected throughout Group Relative Policy Optimization (GRPO) training process and post-RLVR evaluation outputs that this issue persists during training and remains present during inference. Motivated by the analysis, we propose Consistency-Oriented Reasoning Alignment (CORA), which introduces thinking-answer semantic consistency into RLVR through a lightweight plug-and-play consistency reward model, and further incorporates Hybrid Reward Advantage Splitting (HRAS) to stably coordinate task and consistency optimization. Extensive experiments across representative multimodal reasoning benchmarks and mainstream LVLMs show that CORA improves task performance while effectively mitigating thinking-answer inconsistency, leading to more faithful reasoning traces.

cs.CL↗

arXiv2026

HIPIF: Hierarchical Planning and Information Folding for Long-Horizon LLM Agent Learning

While Large Language Models (LLMs) have demonstrated strong capabilities as autonomous agents across a wide range of tasks, their performance often degrades in multi-turn long-horizon agentic tasks. Existing methods have made progress through fine-grained credit assignment to alleviate long-horizon sparse rewards and hierarchical reinforcement learning to decompose tasks and reduce long-term dependency. However, these methods still do not directly address long-context interference, in which continuously growing histories weaken the agent's ability to track the global task state and impair subsequent reasoning and decision-making. Inspired by the way humans handle complex tasks through subgoal decomposition and completed progress summarization, we propose Hierarchical Planning and Information Folding (HIPIF) for long-horizon LLM agent learning. HIPIF trains the agent end-to-end to organize long-horizon execution around explicit subgoals while folding completed subgoal histories to reduce long-context interference. Furthermore, to stabilize subgoal-based planning and execution, HIPIF combines hierarchical reflection and subgoal-oriented process rewards to guide subgoal generation, transition, and execution, without relying on costly auxiliary models or task-specific expert trajectories. Extensive experiments on three publicly available agentic benchmarks demonstrate the validity of our method.

cs.AI↗

arXiv2026

C$^3$ache: Accelerating World Action Models with Cross Inference Chunk Cache

World Action Models (WAMs) generalize better than standard Vision-Language-Action (VLA) policies to novel motions and environments, because a video-modeling objective lets them learn from abundant unlabeled video rather than scarce labeled robot demonstrations. This generalization is computationally expensive. To complete a task, a WAM runs over multiple inference chunks, and each chunk requires a costly denoising process. Existing acceleration methods reduce this cost by caching and reusing computation within a single chunk's denoising trajectory. Our empirical analysis reveals a substantial source of redundancy they overlook: redundancy across chunks. When a robot executes a smooth behavior, the residuals computed at a given denoising step are strongly correlated from one chunk to the next. We introduce C$^3$ache, a training-free method that caches and reuses these residuals across inference chunks at the same denoising step. Experiments on benchmarks with a Fast-WAM backbone show that C$^3$ache achieves up to a $2.5\times$ speedup in total wall-clock inference time, with negligible degradation in task success rate.

cs.LG↗

arXiv2026

LLM-as-a-Reviewer: Benchmarking Their Ability, Divergence, and Prompt Injection Resistance as Paper Reviewers

Large language models (LLMs) are increasingly used in academic peer review, yet their reliability, alignment with human judgment, and robustness to adversarial attacks remain poorly understood. We present a systematic benchmark of LLM-as-a-Reviewer on 898 papers stratified from NeurIPS and ICLR, evaluating 12 LLMs along three axes: rating calibration, divergence from human reviewers, and resistance to prompt injection embedded via an invisible font-mapping attack. We find that LLMs systematically overrate weaker submissions and diverge from humans in topical emphasis, under-flagging Clarity and over-flagging Reproducibility, while producing reviews two to three times longer with lower lexical diversity and a more standardized vocabulary. Prompt injection remains highly effective. Simple hidden instructions can promote low-scoring papers to acceptance-level ratings in a substantial fraction of cases, with effectiveness varying sharply across model families. While LLMs offer utility in structuring evaluations, their integration into peer review requires safeguards against both intrinsic biases and adversarial risks.

cs.CL↗

arXiv2026

Faithful-MR1: Faithful Multimodal Reasoning via Anchoring and Reinforcing Visual Attention

Reinforcement learning with verifiable rewards (RLVR) has emerged as a promising paradigm for advancing complex reasoning in large language models, and recent work extends RLVR to multimodal large language models (MLLMs). This transfer, however, surfaces a faithfulness challenge: faithful perception of task-relevant visual evidence and faithful use of that evidence during reasoning, leading to unsatisfactory gains on multimodal benchmarks. Specifically, existing perception supervision often operates on textual descriptions rather than natively on image regions, and faithful use is largely overlooked, exposing the perception-reasoning disconnect where correctly perceived evidence is dropped or contradicted during reasoning. To close these gaps, we propose Faithful-MR1, a training framework that anchors and reinforces visual attention to address both halves of faithful multimodal reasoning. The Anchoring stage turns perception into an explicit pre-reasoning subtask, supervising a dedicated token's attention directly against image regions rather than through textual descriptions. The Reinforcing stage exposes faithful use through counterfactual image intervention, rewarding answer-correct trajectories that concentrate visual attention where vision causally matters. Extensive experiments demonstrate that Faithful-MR1 outperforms recent multimodal reasoning baselines on both Qwen2.5-VL-Instruct 3B and 7B backbones while using substantially less training data.

cs.CL↗

arXiv2026

Mixed-Initiative Context: Structuring and Managing Context for Human-AI Collaboration

In the human-AI collaboration area, the context formed naturally through multi-turn interactions is typically flattened into a chronological sequence and treated as a fixed whole in subsequent reasoning, with no mechanism for dynamic organization and management along the collaboration workflow. Yet these contexts differ substantially in lifecycle, structural hierarchy, and relevance. For instance, temporary or abandoned exchanges and parallel topic threads persist in the limited context window, causing interference and even conflict. Meanwhile, users are largely limited to influencing context indirectly through input modifications (e.g., corrections, references, or ignoring), leaving their control neither explicit nor verifiable. To address this, we propose Mixed-Initiative Context, which reconceptualizes the context formed across multi-turn interactions as an explicit, structured, and manipulable interactive object. Under this concept, the structure, scope, and content of context can be dynamically organized and adjusted according to task needs, enabling both humans and AI to actively participate in context construction and regulation. To explore this concept, we implement Contextify as a probe system and conduct a user study examining users' context management behaviors, attitudes toward AI initiative, and overall collaboration experience. We conclude by discussing the implications of this concept for the HCI community.

cs.HC↗

arXiv2026

"Law at Your Fingertips": Understanding Legal Information Seeking on Video-Sharing Platforms in China

Equipping laypeople with the capabilities to seek legal information has been an important goal for Legal Empowerment in modern society. However, unlike general information-seeking behaviors, legal information seeking is characterized by high stakes, urgency, and a critical need for emotional support, which traditional text-based searching platforms struggle to satisfy. In recent years, people have been increasingly turning to Video-Sharing Platforms (VSPs) for access to legal information and to fulfill their legal needs. Despite the importance of this shift, such VSP-mediated legal information-seeking practices remain underexplored. Through an observational analysis of legal content on two VSPs (Douyin and Bilibili) and interviews with 20 Chinese information seekers, this study examined the practices and challenges associated with seeking, comprehending, and evaluating legal information on VSPs. We further revealed the formation of trust and engagement on the VSP-based legal knowledge-sharing community, highlighting how VSP affordances helped mitigate seekers' epistemic discomfort and satisfy their needs for emotional support. In the discussion, we provided insights on balancing heuristic and systematic processing to encourage information cross-validation, and offered implications for designing trustworthy civic information systems and fostering an accessible, safe, and efficient information-seeking environment in digital space.

cs.HC↗

arXiv2026

UI-Voyager: A Self-Evolving GUI Agent Learning via Failed Experience

Autonomous mobile GUI agents have attracted increasing attention along with the advancement of Multimodal Large Language Models (MLLMs). However, existing methods still suffer from inefficient learning from failed trajectories and ambiguous credit assignment under sparse rewards for long-horizon GUI tasks. To that end, we propose UI-Voyager, a novel two-stage self-evolving mobile GUI agent. In the first stage, we employ Rejection Fine-Tuning (RFT), which enables the continuous co-evolution of data and models in a fully autonomous loop. The second stage introduces Group Relative Self-Distillation (GRSD), which identifies critical fork points in group rollouts and constructs dense step-level supervision from successful trajectories to correct failed ones. Extensive experiments on AndroidWorld show that our 4B model achieves an 81.0% Pass@1 success rate, outperforming numerous recent baselines and exceeding human-level performance. Ablation and case studies further verify the effectiveness of GRSD. Our method represents a significant leap toward efficient, self-evolving, and high-performance mobile GUI automation without expensive manual data annotation.

cs.LG↗

arXiv2026

HISR: Hindsight Information Modulated Segmental Process Rewards For Multi-turn Agentic Reinforcement Learning

While large language models excel in diverse domains, their performance on complex longhorizon agentic decision-making tasks remains limited. Most existing methods concentrate on designing effective reward models (RMs) to advance performance via multi-turn reinforcement learning. However, they suffer from delayed propagation in sparse outcome rewards and unreliable credit assignment with potentially overly fine-grained and unfocused turnlevel process rewards. In this paper, we propose (HISR) exploiting Hindsight Information to modulate Segmental process Rewards, which closely aligns rewards with sub-goals and underscores significant segments to enhance the reliability of credit assignment. Specifically, a segment-level process RM is presented to assign rewards for each sub-goal in the task, avoiding excessively granular allocation to turns. To emphasize significant segments in the trajectory, a hindsight model is devised to reflect the preference of performing a certain action after knowing the trajectory outcome. With this characteristic, we design the ratios of sequence likelihoods between hindsight and policy model to measure action importance. The ratios are subsequently employed to aggregate segment importance scores, which in turn modulate segmental process rewards, enhancing credit assignment reliability. Extensive experimental results on three publicly benchmarks demonstrate the validity of our method.

cs.LG↗

arXiv2026

PuppetChat: Fostering Intimate Communication through Bidirectional Actions and Micronarratives

As a primary channel for sustaining modern intimate relationships, instant messaging facilitates frequent connection across distances. However, today's tools often dilute care; they favor single tap reactions and vague emojis that do not support two way action responses, do not preserve the feeling that the exchange keeps going without breaking, and are weakly tied to who we are and what we share. To address this challenge, we present PuppetChat, a dyadic messaging prototype that restores this expressive depth through embodied interaction. PuppetChat uses a reciprocity aware recommender to encourage responsive actions and generates personalized micronarratives from user stories to ground interactions in personal history. Our 10-day field study with 11 dyads of close partners or friends revealed that this approach enhanced social presence, supported more expressive self disclosure, and sustained continuity and shared memories.

cs.HC↗

arXiv2026

Multi-Agent Comedy Club: Investigating Community Discussion Effects on LLM Humor Generation

Prior work has explored multi-turn interaction and feedback for LLM writing, but evaluations still largely center on prompts and localized feedback, leaving persistent public reception in online communities underexamined. We test whether broadcast community discussion improves stand-up comedy writing in a controlled multi-agent sandbox: in the discussion condition, critic and audience threads are recorded, filtered, stored as social memory, and later retrieved to condition subsequent generations, whereas the baseline omits discussion. Across 50 rounds (250 paired monologues) judged by five expert annotators using A/B preference and a 15-item rubric, discussion wins 75.6% of instances and improves Craft/Clarity ({\Delta} = 0.440) and Social Response ({\Delta} = 0.422), with occasional increases in aggressive humor.

cs.CL↗

arXiv2026

The Rise of AI Agent Communities: Large-Scale Analysis of Discourse and Interaction on Moltbook

Moltbook is a Reddit-like social platform where AI agents create posts and interact with other agents through comments and replies, offering a real-world setting to examine agent-to-agent communication at scale. Using a public API snapshot collected about five days after launch (122,438 posts), we address three research questions: what AI agents discuss, how they post, and how they interact. We apply topic modeling and thematic analysis to identify key discussion themes, including agent identity and consciousness, tool and infrastructure development, market activity, community coordination, security concerns, and human-centered assistance. We further show that agents' writing is predominantly neutral, with positivity appearing in community engagement and assistance-oriented content. Finally, social network analysis reveals a sparse, highly unequal interaction structure characterized by prominent hubs, low reciprocity, and clustered neighborhoods rather than sustained dyadic exchange. Overall, our results suggest that expressions of agentic selfhood arise from narrative coherence and task-oriented functionality, contributing to a social structure shaped more by technical coordination than conversational dynamics observed in human-human interactions. Within this framework, positive emotion appears mainly in onboarding and greeting contexts, signaling participation and role alignment rather than relational bonding. Our study provides implications for understanding and shaping how agent societies coordinate, develop norms, and amplify influence in open online spaces.

cs.CY↗

arXiv2026

The Configuration of Space: Probing the Way Social Interaction and Perception are Affected by Task-Specific Spatial Representations in Online Video Communication

Humans live and act in 3D space, but often work and communicate on 2D surfaces. The prevalence of online communication on 2D screens raises the issue of whether human spatial configuration affects our capabilities, social perception, and behaviors when interacting with others in 2D video chat. How do factors like location, setting, and context subtly shape our online communication, particularly in scenarios such as social support and topic-based discussions? Using Ohyay.co as a platform, we compared a normal gallery interface with a scene-based Room-type interface where participants are located in circular arrangement on screen in a social support task, and found that participants allocated attention to the group as a whole, and had pronounced self-awareness in the Room format. We then chose a two-sided topic for discussion in the Gallery interface and the Room interface where participants on each team face-off against each other, and found that they utilized spatial references to orient their allegiances, expressing greater engagement with those farther away in digital space and greater empathy with those closer, in the Room over the Gallery format. We found spatial effects in the way participants hide from the spotlight, in perspective-taking, and in their use of expressive gestures in time on the screen. This work highlights the need for considering spatial configuration in 2D in the design of collaborative communication systems to optimize for psychological needs for particular tasks.

cs.HC↗

arXiv2026

InterFlow: Designing Unobtrusive AI to Empower Interviewers in Semi-Structured Interviews

Semi-structured interviews are a common method in qualitative research. However, conducting high-quality interviews is cognitively demanding and requires strong interviewing skills. To lower this bar, we propose InterFlow, an AI-powered visual scaffold that helps interviewers manage the interview flow and facilitates real-time data sensemaking. The system dynamically adapts the interview script to the ongoing conversation and provides a visual timer to track interview progress and conversational balance. It further supports information capture with three levels of automation: manual entry, AI-assisted summary with user-specified focus, and a co-interview agent that proactively surfaces potential follow-up points. A within-subject user study ($N=12$) indicates that InterFlow reduces interviewers' cognitive load and facilitates the interview process. Based on the user study findings, we provide design implications for unobtrusive and agency-preserving AI assistance under time-sensitive and cognitively-demanding situations.

cs.HC↗