arXiv subjects

Zhiqi Shen

Publications and source records attributed to Zhiqi Shen.

At least 37 records · Page 2Linked to original sources

arXiv2026

Following the TRAIL: Predicting and Explaining Tomorrow's Hits with a Fine-Tuned LLM

Large Language Models (LLMs) have been widely applied across multiple domains for their broad knowledge and strong reasoning capabilities. However, applying them to recommendation systems is challenging since it is hard for LLMs to extract user preferences from large, sparse user-item logs, and real-time per-user ranking over the full catalog is too time-consuming to be practical. Moreover, many existing recommender systems focus solely on ranking items while overlooking explanations, which could help improve predictive accuracy and make recommendations more convincing to users. Inspired by recent works that achieve strong recommendation performance by forecasting near-term item popularity, we propose TRAIL (TRend and explAnation Integrated Learner). TRAIL is a fine-tuned LLM that jointly predicts short-term item popularity and generates faithful natural-language explanations. It employs contrastive learning with positive and negative pairs to align its scores and explanations with structured trend signals, yielding accurate and explainable popularity predictions. Extensive experiments show that TRAIL outperforms strong baselines and produces coherent, well-grounded explanations.

cs.IR↗

arXiv2026

FedUTR: Federated Recommendation with Augmented Universal Textual Representation for Sparse Interaction Scenarios

Federated recommendations (FRs) have emerged as an on-device privacy-preserving paradigm, attracting considerable attention driven by rising demands for data security. Existing FRs predominantly adapt ID embeddings to represent items, making the quality of item embeddings entirely dependent on users' historical behaviors. However, we empirically observe that this pattern leads to suboptimal recommendation performance under high data sparsity scenarios, due to its strong reliance on historical interactions. To address this issue, we propose a novel method named FedUTR, which incorporates item textual representations as a complement to interaction behaviors, aiming to enhance model performance under high data sparsity. Specifically, we utilize textual modality as the universal representation to capture generic item knowledge, and design a Collaborative Information Fusion Module (CIFM) to complement each user's personalized interaction information. Besides, we introduce a Local Adaptation Module (LAM) that adaptively exploits the off-the-shelf local model to efficiently preserve client-specific personalized preferences. Moreover, we propose a variant of FedUTR, termed FedUTR-SAR, which incorporates a sparsity-aware resnet component to granularly balance universal and personalized information. The convergence analysis proves theoretical guarantees for the effectiveness of FedUTR. Extensive experiments on four real-world datasets show that our method achieves superior performance, with improvements of up to 59% across all datasets compared to the SOTA baselines.

cs.IR↗

arXiv2026

Towards Comprehensive Stage-wise Benchmarking of Large Language Models in Fact-Checking

Large Language Models (LLMs) are increasingly deployed in real-world fact-checking systems, yet existing evaluations focus predominantly on claim verification and overlook the broader fact-checking workflow, including claim extraction and evidence retrieval. This narrow focus prevents current benchmarks from revealing systematic reasoning failures, factual blind spots, and robustness limitations of modern LLMs. To bridge this gap, we present FactArena, a fully automated arena-style evaluation framework that conducts comprehensive, stage-wise benchmarking of LLMs across the complete fact-checking pipeline. FactArena integrates three key components: (i) an LLM-driven fact-checking process that standardizes claim decomposition, evidence retrieval via tool-augmented interactions, and justification-based verdict prediction; (ii) an arena-styled judgment mechanism guided by consolidated reference guidelines to ensure unbiased and consistent pairwise comparisons across heterogeneous judge agents; and (iii) an arena-driven claim-evolution module that adaptively generates more challenging and semantically controlled claims to probe LLMs' factual robustness beyond fixed seed data. Across 16 state-of-the-art LLMs spanning seven model families, FactArena produces stable and interpretable rankings. Our analyses further reveal significant discrepancies between static claim-verification accuracy and end-to-end fact-checking competence, highlighting the necessity of holistic evaluation. The proposed framework offers a scalable and trustworthy paradigm for diagnosing LLMs' factual reasoning, guiding future model development, and advancing the reliable deployment of LLMs in safety-critical fact-checking applications.

cs.CL↗

arXiv2025

An Efficient Interaction Human-AI Synergy System Bridging Visual Awareness and Large Language Model for Intensive Care Units

Intensive Care Units (ICUs) are critical environments characterized by high-stakes monitoring and complex data management. However, current practices often rely on manual data transcription and fragmented information systems, introducing potential risks to patient safety and operational efficiency. To address these issues, we propose a human-AI synergy system based on a cloud-edge-end architecture, which integrates visual-aware data extraction and semantic interaction mechanisms. Specifically, a visual-aware edge module non-invasively captures real-time physiological data from bedside monitors, reducing manual entry errors. To improve accessibility to fragmented data sources, a semantic interaction module, powered by a Large Language Model (LLM), enables physicians to perform efficient and intuitive voice-based queries over structured patient data. The hierarchical cloud-edge-end deployment ensures low-latency communication and scalable system performance. Our system reduces the cognitive burden on ICU nurses and physicians and demonstrates promising potential for broader applications in intelligent healthcare systems.

cs.HC↗

arXiv2025

Towards Relaxed Multimodal Inputs for Gait-based Parkinson's Disease Assessment

Parkinson's disease assessment has garnered growing interest in recent years, particularly with the advent of sensor data and machine learning techniques. Among these, multimodal approaches have demonstrated strong performance by effectively integrating complementary information from various data sources. However, two major limitations hinder their practical application: (1) the need to synchronize all modalities during training, and (2) the dependence on all modalities during inference. To address these issues, we propose the first Parkinson's assessment system that formulates multimodal learning as a multi-objective optimization (MOO) problem. This not only allows for more flexible modality requirements during both training and inference, but also handles modality collapse issue during multimodal information fusion. In addition, to mitigate the imbalance within individual modalities, we introduce a margin-based class rebalancing strategy to enhance category learning. We conduct extensive experiments on three public datasets under both synchronous and asynchronous settings. The results show that our framework-Towards Relaxed InPuts (TRIP)-achieves state-of-the-art performance, outperforming the best baselines by 16.48, 6.89, and 11.55 percentage points in the asynchronous setting, and by 4.86 and 2.30 percentage points in the synchronous setting, highlighting its effectiveness and adaptability.

cs.AI↗

arXiv2025

Learning an Efficient Optimizer via Hybrid-Policy Sub-Trajectory Balance

Recent advances in generative modeling enable neural networks to generate weights without relying on gradient-based optimization. However, current methods are limited by issues of over-coupling and long-horizon. The former tightly binds weight generation with task-specific objectives, thereby limiting the flexibility of the learned optimizer. The latter leads to inefficiency and low accuracy during inference, caused by the lack of local constraints. In this paper, we propose Lo-Hp, a decoupled two-stage weight generation framework that enhances flexibility through learning various optimization policies. It adopts a hybrid-policy sub-trajectory balance objective, which integrates on-policy and off-policy learning to capture local optimization policies. Theoretically, we demonstrate that learning solely local optimization policies can address the long-horizon issue while enhancing the generation of global optimal weights. In addition, we validate Lo-Hp's superior accuracy and inference efficiency in tasks that require frequent weight updates, such as transfer learning, few-shot learning, domain generalization, and large language model adaptation.

cs.LG↗

arXiv2025

When Retrieval Succeeds and Fails: Rethinking Retrieval-Augmented Generation for LLMs

Large Language Models (LLMs) have enabled a wide range of applications through their powerful capabilities in language understanding and generation. However, as LLMs are trained on static corpora, they face difficulties in addressing rapidly evolving information or domain-specific queries. Retrieval-Augmented Generation (RAG) was developed to overcome this limitation by integrating LLMs with external retrieval mechanisms, allowing them to access up-to-date and contextually relevant knowledge. However, as LLMs themselves continue to advance in scale and capability, the relative advantages of traditional RAG frameworks have become less pronounced and necessary. Here, we present a comprehensive review of RAG, beginning with its overarching objectives and core components. We then analyze the key challenges within RAG, highlighting critical weakness that may limit its effectiveness. Finally, we showcase applications where LLMs alone perform inadequately, but where RAG, when combined with LLMs, can substantially enhance their effectiveness. We hope this work will encourage researchers to reconsider the role of RAG and inspire the development of next-generation RAG systems.

cs.CL↗

arXiv2025

EchoBench: Benchmarking Sycophancy in Medical Large Vision-Language Models

Recent benchmarks for medical Large Vision-Language Models (LVLMs) emphasize leaderboard accuracy, overlooking reliability and safety. We study sycophancy -- models' tendency to uncritically echo user-provided information -- in high-stakes clinical settings. We introduce EchoBench, a benchmark to systematically evaluate sycophancy in medical LVLMs. It contains 2,122 images across 18 departments and 20 modalities with 90 prompts that simulate biased inputs from patients, medical students, and physicians. We evaluate medical-specific, open-source, and proprietary LVLMs. All exhibit substantial sycophancy; the best proprietary model (Claude 3.7 Sonnet) still shows 45.98% sycophancy, and GPT-4.1 reaches 59.15%. Many medical-specific models exceed 95% sycophancy despite only moderate accuracy. Fine-grained analyses by bias type, department, perceptual granularity, and modality identify factors that increase susceptibility. We further show that higher data quality/diversity and stronger domain knowledge reduce sycophancy without harming unbiased accuracy. EchoBench also serves as a testbed for mitigation: simple prompt-level interventions (negative prompting, one-shot, few-shot) produce consistent reductions and motivate training- and decoding-time strategies. Our findings highlight the need for robust evaluation beyond accuracy and provide actionable guidance toward safer, more trustworthy medical LVLMs.

cs.CV↗

arXiv2025

Is Meta-Learning Out? Rethinking Unsupervised Few-Shot Classification with Limited Entropy

Meta-learning is a powerful paradigm for tackling few-shot tasks. However, recent studies indicate that models trained with the whole-class training strategy can achieve comparable performance to those trained with meta-learning in few-shot classification tasks. To demonstrate the value of meta-learning, we establish an entropy-limited supervised setting for fair comparisons. Through both theoretical analysis and experimental validation, we establish that meta-learning has a tighter generalization bound compared to whole-class training. We unravel that meta-learning is more efficient with limited entropy and is more robust to label noise and heterogeneous tasks, making it well-suited for unsupervised tasks. Based on these insights, We propose MINO, a meta-learning framework designed to enhance unsupervised performance. MINO utilizes the adaptive clustering algorithm DBSCAN with a dynamic head for unsupervised task construction and a stability-based meta-scaler for robustness against label noise. Extensive experiments confirm its effectiveness in multiple unsupervised few-shot and zero-shot tasks.

cs.LG↗

arXiv2025

MCP-Universe: Benchmarking Large Language Models with Real-World Model Context Protocol Servers

The Model Context Protocol has emerged as a transformative standard for connecting large language models to external data sources and tools, rapidly gaining adoption across major AI providers and development platforms. However, existing benchmarks are overly simplistic and fail to capture real application challenges such as long-horizon reasoning and large, unfamiliar tool spaces. To address this critical gap, we introduce MCP-Universe, the first comprehensive benchmark specifically designed to evaluate LLMs in realistic and hard tasks through interaction with real-world MCP servers. Our benchmark encompasses 6 core domains spanning 11 different MCP servers: Location Navigation, Repository Management, Financial Analysis, 3D Design, Browser Automation, and Web Searching. To ensure rigorous evaluation, we implement execution-based evaluators, including format evaluators for agent format compliance, static evaluators for time-invariant content matching, and dynamic evaluators that automatically retrieve real-time ground truth for temporally sensitive tasks. Through extensive evaluation of leading LLMs, we find that even SOTA models such as GPT-5 (43.72%), Grok-4 (33.33%) and Claude-4.0-Sonnet (29.44%) exhibit significant performance limitations. In addition, our benchmark poses a significant long-context challenge for LLM agents, as the number of input tokens increases rapidly with the number of interaction steps. Moreover, it introduces an unknown-tools challenge, as LLM agents often lack familiarity with the precise usage of the MCP servers. Notably, enterprise-level agents like Cursor cannot achieve better performance than standard ReAct frameworks. Beyond evaluation, we open-source our extensible evaluation framework with UI support, enabling researchers and practitioners to seamlessly integrate new agents and MCP servers while fostering innovation in the rapidly evolving MCP ecosystem.

cs.AI↗

arXiv2025

Modeling Human Responses to Multimodal AI Content

As AI-generated content becomes widespread, so does the risk of misinformation. While prior research has primarily focused on identifying whether content is authentic, much less is known about how such content influences human perception and behavior. In domains like trading or the stock market, predicting how people react (e.g., whether a news post will go viral), can be more critical than verifying its factual accuracy. To address this, we take a human-centered approach and introduce the MhAIM Dataset, which contains 154,552 online posts (111,153 of them AI-generated), enabling large-scale analysis of how people respond to AI-generated content. Our human study reveals that people are better at identifying AI content when posts include both text and visuals, particularly when inconsistencies exist between the two. We propose three new metrics: trustworthiness, impact, and openness, to quantify how users judge and engage with online content. We present T-Lens, an LLM-based agent system designed to answer user queries by incorporating predicted human responses to multimodal information. At its core is HR-MCP (Human Response Model Context Protocol), built on the standardized Model Context Protocol (MCP), enabling seamless integration with any LLM. This integration allows T-Lens to better align with human reactions, enhancing both interpretability and interaction capabilities. Our work provides empirical insights and practical tools to equip LLMs with human-awareness capabilities. By highlighting the complex interplay among AI, human cognition, and information reception, our findings suggest actionable strategies for mitigating the risks of AI-driven misinformation.

cs.AI↗

arXiv2025

Semantic Item Graph Enhancement for Multimodal Recommendation

Multimodal recommendation systems have attracted increasing attention for their improved performance by leveraging items' multimodal information. Prior methods often build modality-specific item-item semantic graphs from raw modality features and use them as supplementary structures alongside the user-item interaction graph to enhance user preference learning. However, these semantic graphs suffer from semantic deficiencies, including (1) insufficient modeling of collaborative signals among items and (2) structural distortions introduced by noise in raw modality features, ultimately compromising performance. To address these issues, we first extract collaborative signals from the interaction graph and infuse them into each modality-specific item semantic graph to enhance semantic modeling. Then, we design a modulus-based personalized embedding perturbation mechanism that injects perturbations with modulus-guided personalized intensity into embeddings to generate contrastive views. This enables the model to learn noise-robust representations through contrastive learning, thereby reducing the effect of structural noise in semantic graphs. Besides, we propose a dual representation alignment mechanism that first aligns multiple semantic representations via a designed Anchor-based InfoNCE loss using behavior representations as anchors, and then aligns behavior representations with the fused semantics by standard InfoNCE, to ensure representation consistency. Extensive experiments on four benchmark datasets validate the effectiveness of our framework.

cs.IR↗

arXiv2025

Does Multimodality Improve Recommender Systems as Expected? A Critical Analysis and Future Directions

Multimodal recommendation systems are increasingly popular for their potential to improve performance by integrating diverse data types. However, the actual benefits of this integration remain unclear, raising questions about when and how it truly enhances recommendations. In this paper, we propose a structured evaluation framework to systematically assess multimodal recommendations across four dimensions: Comparative Efficiency, Recommendation Tasks, Recommendation Stages, and Multimodal Data Integration. We benchmark a set of reproducible multimodal models against strong traditional baselines and evaluate their performance on different platforms. Our findings show that multimodal data is particularly beneficial in sparse interaction scenarios and during the recall stage of recommendation pipelines. We also observe that the importance of each modality is task-specific, where text features are more useful in e-commerce and visual features are more effective in short-video recommendations. Additionally, we explore different integration strategies and model sizes, finding that Ensemble-Based Learning outperforms Fusion-Based Learning, and that larger models do not necessarily deliver better results. To deepen our understanding, we include case studies and review findings from other recommendation domains. Our work provides practical insights for building efficient and effective multimodal recommendation systems, emphasizing the need for thoughtful modality selection, integration strategies, and model design.

cs.IR↗

arXiv2025

CM$^3$: Calibrating Multimodal Recommendation

Alignment and uniformity are fundamental principles within the domain of contrastive learning. In recommender systems, prior work has established that optimizing the Bayesian Personalized Ranking (BPR) loss contributes to the objectives of alignment and uniformity. Specifically, alignment aims to draw together the representations of interacting users and items, while uniformity mandates a uniform distribution of user and item embeddings across a unit hypersphere. This study revisits the alignment and uniformity properties within the context of multimodal recommender systems, revealing a proclivity among extant models to prioritize uniformity to the detriment of alignment. Our hypothesis challenges the conventional assumption of equitable item treatment through a uniformity loss, proposing a more nuanced approach wherein items with similar multimodal attributes converge toward proximal representations within the hyperspheric manifold. Specifically, we leverage the inherent similarity between items' multimodal data to calibrate their uniformity distribution, thereby inducing a more pronounced repulsive force between dissimilar entities within the embedding space. A theoretical analysis elucidates the relationship between this calibrated uniformity loss and the conventional uniformity function. Moreover, to enhance the fusion of multimodal features, we introduce a Spherical Bézier method designed to integrate an arbitrary number of modalities while ensuring that the resulting fused features are constrained to the same hyperspherical manifold. Empirical evaluations conducted on five real-world datasets substantiate the superiority of our approach over competing baselines. We also shown that the proposed methods can achieve up to a 5.4% increase in NDCG@20 performance via the integration of MLLM-extracted features. Source code is available at: https://github.com/enoche/CM3.

cs.IR↗

arXiv2025

RoleRAG: Enhancing LLM Role-Playing via Graph Guided Retrieval

Large Language Models (LLMs) have shown promise in character imitation, enabling immersive and engaging conversations. However, they often generate content that is irrelevant or inconsistent with a character's background. We attribute these failures to: (1) the inability to accurately recall character-specific knowledge due to entity ambiguity, and (2) a lack of awareness of the character's cognitive boundaries. To address these issues, we propose RoleRAG, a retrieval-based framework that integrates efficient entity disambiguation for knowledge indexing with a boundary-aware retriever for extracting contextually appropriate information from a structured knowledge graph. Experiments on role-playing benchmarks show that RoleRAG's calibrated retrieval helps both general-purpose and role-specific LLMs better align with character knowledge and reduce hallucinated responses.

cs.AI↗

arXiv2025

Response Uncertainty and Probe Modeling: Two Sides of the Same Coin in LLM Interpretability?

Probing techniques have shown promise in revealing how LLMs encode human-interpretable concepts, particularly when applied to curated datasets. However, the factors governing a dataset's suitability for effective probe training are not well-understood. This study hypothesizes that probe performance on such datasets reflects characteristics of both the LLM's generated responses and its internal feature space. Through quantitative analysis of probe performance and LLM response uncertainty across a series of tasks, we find a strong correlation: improved probe performance consistently corresponds to a reduction in response uncertainty, and vice versa. Subsequently, we delve deeper into this correlation through the lens of feature importance analysis. Our findings indicate that high LLM response variance is associated with a larger set of important features, which poses a greater challenge for probe models and often results in diminished performance. Moreover, leveraging the insights from response uncertainty analysis, we are able to identify concrete examples where LLM representations align with human knowledge across diverse domains, offering additional evidence of interpretable reasoning in LLMs.

cs.AI↗

arXiv2025

GiFT: Gibbs Fine-Tuning for Code Generation

Training Large Language Models (LLMs) with synthetic data is a prevalent practice in code generation. A key approach is self-training, where LLMs are iteratively trained on self-generated correct code snippets. In this case, the self-generated codes are drawn from a conditional distribution, conditioned on a specific seed description. However, the seed description is not the only valid representation that aligns with its intended meaning. With all valid descriptions and codes forming a joint space, codes drawn from the conditional distribution would lead to an underrepresentation of the full description-code space. As such, we propose Gibbs Fine-Tuning (GiFT), a novel self-training method inspired by Gibbs sampling. GiFT allows self-generated data to be drawn from the marginal distribution of the joint space, thereby mitigating the biases inherent in conditional sampling. We provide a theoretical analysis demonstrating the potential benefits of fine-tuning LLMs with code derived from the marginal distribution. Furthermore, we propose a perplexity-based code selection method to mitigate the imbalanced long-tail distribution of the self-generated codes. Empirical evaluation of two LLMs across four datasets demonstrates that GiFT achieves superior performance, particularly on more challenging benchmarks. Source code is available at https://github.com/Alex-HaochenLi/GiFT.

cs.LG↗

arXiv2025

Learning to Learn Weight Generation via Local Consistency Diffusion

Diffusion-based algorithms have emerged as promising techniques for weight generation. However, existing solutions are limited by two challenges: generalizability and local target assignment. The former arises from the inherent lack of cross-task transferability in existing single-level optimization methods, limiting the model's performance on new tasks. The latter lies in existing research modeling only global optimal weights, neglecting the supervision signals in local target weights. Moreover, naively assigning local target weights causes local-global inconsistency. To address these issues, we propose Mc-Di, which integrates the diffusion algorithm with meta-learning for better generalizability. Furthermore, we extend the vanilla diffusion into a local consistency diffusion algorithm. Our theory and experiments demonstrate that it can learn from local targets while maintaining consistency with the global optima. We validate Mc-Di's superior accuracy and inference efficiency in tasks that require frequent weight updates, including transfer learning, few-shot learning, domain generalization, and large language model adaptation.

cs.LG↗