arXiv subjects

Ziwei Zhu

Publications and source records attributed to Ziwei Zhu.

At least 37 records · Page 2Linked to original sources

arXiv2026

Context-Aware Decoding for Faithful Vision-Language Generation

Hallucinations, generating responses inconsistent with the visual input, remain a critical limitation of large vision-language models (LVLMs), especially in open-ended tasks such as image captioning and visual reasoning. In this work, we probe the layer-wise generation dynamics that drive hallucinations and propose a training-free mitigation strategy. Employing the Logit Lens, we examine how LVLMs construct next-token distributions across decoder layers, uncovering a pronounced commitment-depth gap: truthful tokens accumulate probability mass on their final candidates earlier than hallucinatory ones. Drawing on this discovery, we introduce Context Embedding Injection (CEI), a lightweight method that harnesses the hidden state of the last input token-the context embedding-as a grounding signal to maintain visual fidelity throughout decoding and curb hallucinations. Evaluated on the CHAIR, AMBER, and MMHal-Bench benchmarks (with a maximum token length of 512), CEI outperforms state-of-the-art baselines across three LVLMs, with its dynamic variant yielding the lowest overall hallucination rates. By integrating novel mechanistic insights with a scalable intervention, this work advances the mitigation of hallucinations in LVLMs.

cs.CV↗

arXiv2025

ORGEval: Graph-Theoretic Evaluation of LLMs in Optimization Modeling

Formulating optimization problems for industrial applications demands significant manual effort and domain expertise. While Large Language Models (LLMs) show promise in automating this process, evaluating their performance remains difficult due to the absence of robust metrics. Existing solver-based approaches often face inconsistency, infeasibility issues, and high computational costs. To address these issues, we propose ORGEval, a graph-theoretic evaluation framework for assessing LLMs' capabilities in formulating linear and mixed-integer linear programs. ORGEval represents optimization models as graphs, reducing equivalence detection to graph isomorphism testing. We identify and prove a sufficient condition, when the tested graphs are symmetric decomposable (SD), under which the Weisfeiler-Lehman (WL) test is guaranteed to correctly detect isomorphism. Building on this, ORGEval integrates a tailored variant of the WL-test with an SD detection algorithm to evaluate model equivalence. By focusing on structural equivalence rather than instance-level configurations, ORGEval is robust to numerical variations. Experimental results show that our method can successfully detect model equivalence and produce 100\% consistent results across random parameter configurations, while significantly outperforming solver-based methods in runtime, especially on difficult problems. Leveraging ORGEval, we construct the Bench4Opt dataset and benchmark state-of-the-art LLMs on optimization modeling. Our results reveal that although optimization modeling remains challenging for all LLMs, DeepSeek-V3 and Claude-Opus-4 achieve the highest accuracies under direct prompting, outperforming even leading reasoning models.

cs.LG↗

arXiv2025

ChronoPlay: A Framework for Modeling Dual Dynamics and Authenticity in Game RAG Benchmarks

Retrieval Augmented Generation (RAG) systems are increasingly vital in dynamic domains like online gaming, yet the lack of a dedicated benchmark has impeded standardized evaluation in this area. The core difficulty lies in Dual Dynamics: the constant interplay between game content updates and the shifting focus of the player community. Furthermore, the necessity of automating such a benchmark introduces a critical requirement for player-centric authenticity to ensure generated questions are realistic. To address this integrated challenge, we introduce ChronoPlay, a novel framework for the automated and continuous generation of game RAG benchmarks. ChronoPlay utilizes a dual-dynamic update mechanism to track both forms of change, and a dual-source synthesis engine that draws from official sources and player community to ensure both factual correctness and authentic query patterns. We instantiate our framework on three distinct games to create the first dynamic RAG benchmark for the gaming domain, offering new insights into model performance under these complex and realistic conditions. Code is avaliable at: https://github.com/hly1998/ChronoPlay.

cs.CL↗

arXiv2025

CORTEX: Collaborative LLM Agents for High-Stakes Alert Triage

Security Operations Centers (SOCs) are overwhelmed by tens of thousands of daily alerts, with only a small fraction corresponding to genuine attacks. This overload creates alert fatigue, leading to overlooked threats and analyst burnout. Classical detection pipelines are brittle and context-poor, while recent LLM-based approaches typically rely on a single model to interpret logs, retrieve context, and adjudicate alerts end-to-end -- an approach that struggles with noisy enterprise data and offers limited transparency. We propose CORTEX, a multi-agent LLM architecture for high-stakes alert triage in which specialized agents collaborate over real evidence: a behavior-analysis agent inspects activity sequences, evidence-gathering agents query external systems, and a reasoning agent synthesizes findings into an auditable decision. To support training and evaluation, we release a dataset of fine-grained SOC investigations from production environments, capturing step-by-step analyst actions and linked tool outputs. Across diverse enterprise scenarios, CORTEX substantially reduces false positives and improves investigation quality over state-of-the-art single-agent LLMs.

cs.CL↗

arXiv2025

What's Not Said Still Hurts: A Description-Based Evaluation Framework for Measuring Social Bias in LLMs

Large Language Models (LLMs) often exhibit social biases inherited from their training data. While existing benchmarks evaluate bias by term-based mode through direct term associations between demographic terms and bias terms, LLMs have become increasingly adept at avoiding biased responses, leading to seemingly low levels of bias. However, biases persist in subtler, contextually hidden forms that traditional benchmarks fail to capture. We introduce the Description-based Bias Benchmark (DBB), a novel dataset designed to assess bias at the semantic level that bias concepts are hidden within naturalistic, subtly framed contexts in real-world scenarios rather than superficial terms. We analyze six state-of-the-art LLMs, revealing that while models reduce bias in response at the term level, they continue to reinforce biases in nuanced settings. Data, code, and results are available at https://github.com/JP-25/Description-based-Bias-Benchmark.

cs.CL↗

arXiv2025

Mitigating Hallucination in Large Vision-Language Models via Adaptive Attention Calibration

Large vision-language models (LVLMs) achieve impressive performance on multimodal tasks but often suffer from hallucination, and confidently describe objects or attributes not present in the image. Current training-free interventions struggle to maintain accuracy in open-ended and long-form generation scenarios. We introduce the Confidence-Aware Attention Calibration (CAAC) framework to address this challenge by targeting two key biases: spatial perception bias, which distributes attention disproportionately across image tokens, and modality bias, which shifts focus from visual to textual inputs over time. CAAC employs a two-step approach: Visual-Token Calibration (VTC) to balance attention across visual tokens, and Adaptive Attention Re-Scaling (AAR) to reinforce visual grounding guided by the model's confidence. This confidence-driven adjustment ensures consistent visual alignment during generation. Experiments on CHAIR, AMBER, and POPE benchmarks demonstrate that CAAC outperforms baselines, particularly in long-form generations, effectively reducing hallucination.

cs.CV↗

arXiv2025

Learning to Explain: Prototype-Based Surrogate Models for LLM Classification

Large language models (LLMs) have demonstrated impressive performance on natural language tasks, but their decision-making processes remain largely opaque. Existing explanation methods either suffer from limited faithfulness to the model's reasoning or produce explanations that humans find difficult to understand. To address these challenges, we propose \textbf{ProtoSurE}, a novel prototype-based surrogate framework that provides faithful and human-understandable explanations for LLMs. ProtoSurE trains an interpretable-by-design surrogate model that aligns with the target LLM while utilizing sentence-level prototypes as human-understandable concepts. Extensive experiments show that ProtoSurE consistently outperforms SOTA explanation methods across diverse LLMs and datasets. Importantly, ProtoSurE demonstrates strong data efficiency, requiring relatively few training examples to achieve good performance, making it practical for real-world applications.

cs.CL↗

arXiv2025

Spurious Correlations and Beyond: Understanding and Mitigating Shortcut Learning in SDOH Extraction with Large Language Models

Social determinants of health (SDOH) extraction from clinical text is critical for downstream healthcare analytics. Although large language models (LLMs) have shown promise, they may rely on superficial cues leading to spurious predictions. Using the MIMIC portion of the SHAC (Social History Annotation Corpus) dataset and focusing on drug status extraction as a case study, we demonstrate that mentions of alcohol or smoking can falsely induce models to predict current/past drug use where none is present, while also uncovering concerning gender disparities in model performance. We further evaluate mitigation strategies - such as prompt engineering and chain-of-thought reasoning - to reduce these false positives, providing insights into enhancing LLM reliability in health domains.

cs.CL↗

arXiv2025

Understanding Best Subset Selection: A Tale of Two C(omplex)ities

We consider the problem of best subset selection (BSS) under high-dimensional sparse linear regression model. Recently, Guo et al. (2020) showed that the model selection performance of BSS depends on a certain identifiability margin, a measure that captures the model discriminative power of BSS under a general correlation structure that is robust to the design dependence, unlike its computational surrogates such as LASSO, SCAD, MCP, etc. Expanding on this, we further broaden the theoretical understanding of best subset selection in this paper and show that the complexities of the residualized signals, the portion of the signals orthogonal to the true active features, and spurious projections, describing the projection operators associated with the irrelevant features, also play fundamental roles in characterizing the margin condition for model consistency of BSS. In particular, we establish both necessary and sufficient margin conditions depending only on the identifiability margin and the two complexity measures. We also partially extend our sufficiency result to the case of high-dimensional sparse generalized linear models (GLMs).

math.ST↗

arXiv2025

Federated Conversational Recommender System

Conversational Recommender Systems (CRSs) have become increasingly popular as a powerful tool for providing personalized recommendation experiences. By directly engaging with users in a conversational manner to learn their current and fine-grained preferences, a CRS can quickly derive recommendations that are relevant and justifiable. However, existing conversational recommendation systems (CRSs) typically rely on a centralized training and deployment process, which involves collecting and storing explicitly-communicated user preferences in a centralized repository. These fine-grained user preferences are completely human-interpretable and can easily be used to infer sensitive information (e.g., financial status, political stands, and health information) about the user, if leaked or breached. To address the user privacy concerns in CRS, we first define a set of privacy protection guidelines for preserving user privacy under the conversational recommendation setting. Based on these guidelines, we propose a novel federated conversational recommendation framework that effectively reduces the risk of exposing user privacy by (i) de-centralizing both the historical interests estimation stage and the interactive preference elicitation stage and (ii) strictly bounding privacy leakage by enforcing user-level differential privacy with meticulously selected privacy budgets. Through extensive experiments, we show that the proposed framework not only satisfies these user privacy protection guidelines, but also enables the system to achieve competitive recommendation performance even when compared to the state-of-the-art non-private conversational recommendation approach.

cs.IR↗

arXiv2025

Fighting Spurious Correlations in Text Classification via a Causal Learning Perspective

In text classification tasks, models often rely on spurious correlations for predictions, incorrectly associating irrelevant features with the target labels. This issue limits the robustness and generalization of models, especially when faced with out-of-distribution data where such spurious correlations no longer hold. To address this challenge, we propose the Causally Calibrated Robust Classifier (CCR), which aims to reduce models' reliance on spurious correlations and improve model robustness. Our approach integrates a causal feature selection method based on counterfactual reasoning, along with an unbiased inverse propensity weighting (IPW) loss function. By focusing on selecting causal features, we ensure that the model relies less on spurious features during prediction. We theoretically justify our approach and empirically show that CCR achieves state-of-the-art performance among methods without group labels, and in some cases, it can compete with the models that utilize group labels.

cs.LG↗

arXiv2024

Generative Semantic Communication: Architectures, Technologies, and Applications

This paper delves into the applications of generative artificial intelligence (GAI) in semantic communication (SemCom) and presents a thorough study. Three popular SemCom systems enabled by classical GAI models are first introduced, including variational autoencoders, generative adversarial networks, and diffusion models. For each system, the fundamental concept of the GAI model, the corresponding SemCom architecture, and the associated literature review of recent efforts are elucidated. Then, a novel generative SemCom system is proposed by incorporating the cutting-edge GAI technology-large language models (LLMs). This system features two LLM-based AI agents at both the transmitter and receiver, serving as "brains" to enable powerful information understanding and content regeneration capabilities, respectively. This innovative design allows the receiver to directly generate the desired content, instead of recovering the bit stream, based on the coded semantic information conveyed by the transmitter. Therefore, it shifts the communication mindset from "information recovery" to "information regeneration" and thus ushers in a new era of generative SemCom. A case study on point-to-point video retrieval is presented to demonstrate the superiority of the proposed generative SemCom system, showcasing a 99.98% reduction in communication overhead and a 53% improvement in retrieval accuracy compared to the traditional communication system. Furthermore, four typical application scenarios for generative SemCom are delineated, followed by a discussion of three open issues warranting future investigation. In a nutshell, this paper provides a holistic set of guidelines for applying GAI in SemCom, paving the way for the efficient implementation of generative SemCom in future wireless networks.

cs.IT↗

arXiv2024

Cross-Domain Recommendation Meets Large Language Models

Cross-domain recommendation (CDR) has emerged as a promising solution to the cold-start problem, faced by single-domain recommender systems. However, existing CDR models rely on complex neural architectures, large datasets, and significant computational resources, making them less effective in data-scarce scenarios or when simplicity is crucial. In this work, we leverage the reasoning capabilities of large language models (LLMs) and explore their performance in the CDR domain across multiple domain pairs. We introduce two novel prompt designs tailored for CDR and demonstrate that LLMs, when prompted effectively, outperform state-of-the-art CDR baselines across various metrics and domain combinations in the rating prediction and ranking tasks. This work bridges the gap between LLMs and recommendation systems, showcasing their potential as effective cross-domain recommenders.

cs.IR↗

arXiv2024

ORIS: Online Active Learning Using Reinforcement Learning-based Inclusive Sampling for Robust Streaming Analytics System

Effective labeled data collection plays a critical role in developing and fine-tuning robust streaming analytics systems. However, continuously labeling documents to filter relevant information poses significant challenges like limited labeling budget or lack of high-quality labels. There is a need for efficient human-in-the-loop machine learning (HITL-ML) design to improve streaming analytics systems. One particular HITL- ML approach is online active learning, which involves iteratively selecting a small set of the most informative documents for labeling to enhance the ML model performance. The performance of such algorithms can get affected due to human errors in labeling. To address these challenges, we propose ORIS, a method to perform Online active learning using Reinforcement learning-based Inclusive Sampling of documents for labeling. ORIS aims to create a novel Deep Q-Network-based strategy to sample incoming documents that minimize human errors in labeling and enhance the ML model performance. We evaluate the ORIS method on emotion recognition tasks, and it outperforms traditional baselines in terms of both human labeling performance and the ML model performance.

cs.LG↗

arXiv2024

Advancing Interpretability in Text Classification through Prototype Learning

Deep neural networks have achieved remarkable performance in various text-based tasks but often lack interpretability, making them less suitable for applications where transparency is critical. To address this, we propose ProtoLens, a novel prototype-based model that provides fine-grained, sub-sentence level interpretability for text classification. ProtoLens uses a Prototype-aware Span Extraction module to identify relevant text spans associated with learned prototypes and a Prototype Alignment mechanism to ensure prototypes are semantically meaningful throughout training. By aligning the prototype embeddings with human-understandable examples, ProtoLens provides interpretable predictions while maintaining competitive accuracy. Extensive experiments demonstrate that ProtoLens outperforms both prototype-based and non-interpretable baselines on multiple text classification benchmarks. Code and data are available at \url{https://anonymous.4open.science/r/ProtoLens-CE0B/}.

cs.CL↗

arXiv2024

Crossroads of Continents: Automated Artifact Extraction for Cultural Adaptation with Large Multimodal Models

We present a comprehensive three-phase study to examine (1) the cultural understanding of Large Multimodal Models (LMMs) by introducing DalleStreet, a large-scale dataset generated by DALL-E 3 and validated by humans, containing 9,935 images of 67 countries and 10 concept classes; (2) the underlying implicit and potentially stereotypical cultural associations with a cultural artifact extraction task; and (3) an approach to adapt cultural representation in an image based on extracted associations using a modular pipeline, CultureAdapt. We find disparities in cultural understanding at geographic sub-region levels with both open-source (LLaVA) and closed-source (GPT-4V) models on DalleStreet and other existing benchmarks, which we try to understand using over 18,000 artifacts that we identify in association to different countries. Our findings reveal a nuanced picture of the cultural competence of LMMs, highlighting the need to develop culture-aware systems. Dataset and code are available at https://github.com/iamshnoo/crossroads

cs.CL↗

arXiv2024

BiasDora: Exploring Hidden Biased Associations in Vision-Language Models

Existing works examining Vision-Language Models (VLMs) for social biases predominantly focus on a limited set of documented bias associations, such as gender:profession or race:crime. This narrow scope often overlooks a vast range of unexamined implicit associations, restricting the identification and, hence, mitigation of such biases. We address this gap by probing VLMs to (1) uncover hidden, implicit associations across 9 bias dimensions. We systematically explore diverse input and output modalities and (2) demonstrate how biased associations vary in their negativity, toxicity, and extremity. Our work (3) identifies subtle and extreme biases that are typically not recognized by existing methodologies. We make the Dataset of retrieved associations, (Dora), publicly available here https://github.com/chahatraj/BiasDora.

cs.CL↗

arXiv2024

Navigating the Shortcut Maze: A Comprehensive Analysis of Shortcut Learning in Text Classification by Language Models

Language models (LMs), despite their advances, often depend on spurious correlations, undermining their accuracy and generalizability. This study addresses the overlooked impact of subtler, more complex shortcuts that compromise model reliability beyond oversimplified shortcuts. We introduce a comprehensive benchmark that categorizes shortcuts into occurrence, style, and concept, aiming to explore the nuanced ways in which these shortcuts influence the performance of LMs. Through extensive experiments across traditional LMs, large language models, and state-of-the-art robust models, our research systematically investigates models' resilience and susceptibilities to sophisticated shortcuts. Our benchmark and code can be found at: https://github.com/yuqing-zhou/shortcut-learning-in-text-classification.

cs.CL↗