arXiv subjects

Ee-Peng Lim

Publications and source records attributed to Ee-Peng Lim.

At least 19 recordsLinked to original sources

arXiv2026

Do AI Personas Grow? Analyzing and Benchmarking Personality Evolution in LLM Agents After Life Events

Personality-conditioned LLM agents (PC-Agents) are increasingly used in emotional support, social simulation, and role-playing, motivating the development of lifelong agents that remain coherent over extended interactions. A key component of such coherence is personality evolution: agents should undergo plausible, psychology-grounded changes as they experience life events in different contexts. Although prior work shows that LLM personalities can shift under contextual perturbations, how these shifts vary across traits, events, personas, and models remains poorly understood. We study event-induced personality change after 11 major life events, using the Big Five traits as a psychometric anchor and interpreting the resulting trajectories against longitudinal evidence from human personality psychology. Across four diagnostic axes, PC-Agents exhibit measurable trait shifts at similar rates for event-trait pairs with and without documented human change directions. Even when shifts follow the expected direction, their magnitudes usually fall below human effect-size ranges. Gender and cultural-region prompts show little moderating effect, while persona-level dispersion is compressed three- to four-fold relative to human samples. To enable systematic comparison, we introduce BFI-Adapt, a reusable benchmark for scoring the directional fidelity of event-induced personality change, and use it to rank 14 models. A validation suite shows that the measured shifts exceed no-event retest noise, remain stable under independently paraphrased prompts, exhibit limited and model-dependent convergence with scenario-based behavioral choices, and persist across intervening unrelated dialogue. Together, these checks establish the measured trajectories as robust event-conditioned response patterns. Our results suggest that current PC-Agents simulate the mean of human personality dynamics, but not its shape.

cs.CL↗

arXiv2026

MIThinker: A Plug-and-Play Policy-Optimized Thinker For Motivational Interviewing Counseling

Reasoning large language models (LLMs) have recently made much progress in complex problem-solving, leveraging internal reasoning (or thought) to guide their solution generation. However, existing LLM-based counseling agents, including those using Motivational Interviewing (MI), generate responses without explicitly aligning thoughts with counseling techniques, limiting their effectiveness. We propose MIThinker, a lightweight thinking model that generates therapeutic thoughts to guide MI counseling agents in strategy selection and response generation. To overcome the lack of annotated thought data, we introduce AugR1-MI, an automated pipeline that reverse-engineers counselor's thoughts from observed responses. Through two-stage training combining supervised fine-tuning and reinforcement learning, MIThinker demonstrates improved theory-of-mind assessment and strategy alignment. Comprehensive evaluations show that MindfulMI, our agent leveraging MIThinker, achieves MI competency comparable to state-of-the-art systems with an order of magnitude less computation.

cs.CL↗

arXiv2026

VerifyMAS: Hypothesis Verification for Failure Attribution in LLM Multi-Agent Systems

Large language model-driven multi-agent systems (LLM-MAS) excel at complex tasks, yet unreliable agents remain a key bottleneck to system-level reliability. Automatic failure attribution is therefore critical, but existing approaches, such as direct prediction of agent-error pairs and agent-first failure attribution, rely on local logs of agents and miss global failures that only manifest over full interaction trajectories, such as cross-step inconsistencies and inter-agent coordination errors. Moreover, directly predicting failures induces a large combinatorial search space, hindering fine-grained attribution. To address these challenges, we propose VerifyMAS, a hypothesis verification framework for agent failure attribution. Instead of directly predicting faulty agents and error types, VerifyMAS formulates and verifies failure hypotheses against full trajectories. This verification-based approach decomposes attribution into trajectory-level error validation and fine-grained agent localization, providing an error-first attribution approach that captures global failure patterns while substantially reducing the search space. We further introduce a hypothesis-based data construction strategy grounded in a structured error taxonomy and fine-tune a specialized LLM verifier model for trajectory-level failure verification and agent attribution. Experiments on Aegis-Bench and Who&When show that VerifyMAS consistently improves diverse backbone models, including open-source Qwen and API-based GPT models, outperforming prior methods without sacrificing inference efficiency for long multi-agent trajectories.

cs.CL↗

arXiv2026

On Reasoning Behind Next Occupation Recommendation

In this work, we develop a novel reasoning approach to enhance the performance of large language models (LLMs) in future occupation prediction. In this approach, a reason generator first derives a ``reason'' for a user using his/her past education and career history. The reason summarizes the user's preference and is used as the input of an occupation predictor to recommend the user's next occupation. This two-step occupation prediction approach is, however, non-trivial as LLMs are not aligned with career paths or the unobserved reasons behind each occupation decision. We therefore propose to fine-tune LLMs improving their reasoning and occupation prediction performance. We first derive high-quality oracle reasons, as measured by factuality, coherence and utility criteria, using a LLM-as-a-Judge. These oracle reasons are then used to fine-tune small LLMs to perform reason generation and next occupation prediction. Our extensive experiments show that: (a) our approach effectively enhances LLM's accuracy in next occupation prediction making them comparable to fully supervised methods and outperforming unsupervised methods; (b) a single LLM fine-tuned to perform reason generation and occupation prediction outperforms two LLMs fine-tuned to perform the tasks separately; and (c) the next occupation prediction accuracy depends on the quality of generated reasons. Our code is available at https://github.com/Sarasarahhhhh/job_prediction.

cs.CL↗

arXiv2026

DialToM: A Theory of Mind Benchmark for Forecasting State-Driven Dialogue Trajectories

We introduce DialToM, an annotated Theory of Mind (ToM) benchmark built from naturalistic human-human dialogues using a multiple-choice evaluation framework. Concurrent with recent work showing a gap between explicit mental-state inference and applied ToM in synthetic settings~\cite{gu2024simpletom}, we establish a stricter \emph{State-Driven Diagnostic Probe} in which models must forecast state-consistent dialogue trajectories solely from isolated mental-state profiles without dialogue context. Our evaluation reveals a systematic reasoning asymmetry -- LLMs excel at inferring mental states (Literal ToM) but struggle to leverage them for social forecasting (Functional ToM). Crucially, a domain expert achieves 100\% accuracy on this task, proving its validity and establishing a stark human-AI capability gap. Further, a teacher-student reasoning injection probe shows that Gemini 3 Pro -- which establishes the leading baseline -- possesses robust Functional ToM capabilities for context-free forecasting that are transferable to weaker models. DialToM, its evaluation code, and dataset are publicly available at https://github.com/Stealth-py/DialToM.

cs.CL↗

arXiv2026

MHSafeEval: Role-Aware Interaction-Level Evaluation of Mental Health Safety in Large Language Models

Large language models (LLMs) are increasingly explored as scalable tools for mental health counseling, yet evaluating their safety remains challenging due to the interactional and context-dependent nature of clinical harm. Existing evaluation frameworks predominantly assess isolated responses using coarse-grained taxonomies or static datasets, limiting their ability to diagnose how harms emerge and accumulate over multi-turn counseling interactions. In this work, we introduce R-MHSafe, a role-aware mental health safety taxonomy that characterizes clinically significant harm in terms of the interactional roles an AI counselor adopts, including perpetrator, instigator, facilitator, or enabler, combined with clinically grounded harm categories. Then, we propose MHSafeEval, a closed-loop, agent-based evaluation framework that formulates safety assessment as trajectory-level discovery of harm through adversarial multi-turn interactions, guided by role-aware modeling. Using R-MHSafe and MHSafeEval, we conduct a large-scale evaluation across state-of-the-art LLMs. Our results reveal substantial role-dependent and cumulative safety failures that are systematically missed by existing static benchmarks, and show that our framework significantly improves failure-mode coverage and diagnostic granularity.

cs.CL↗

arXiv2025

LLMs-based Augmentation for Domain Adaptation in Long-tailed Food Datasets

Training a model for food recognition is challenging because the training samples, which are typically crawled from the Internet, are visually different from the pictures captured by users in the free-living environment. In addition to this domain-shift problem, the real-world food datasets tend to be long-tailed distributed and some dishes of different categories exhibit subtle variations that are difficult to distinguish visually. In this paper, we present a framework empowered with large language models (LLMs) to address these challenges in food recognition. We first leverage LLMs to parse food images to generate food titles and ingredients. Then, we project the generated texts and food images from different domains to a shared embedding space to maximize the pair similarities. Finally, we take the aligned features of both modalities for recognition. With this simple framework, we show that our proposed approach can outperform the existing approaches tailored for long-tailed data distribution, domain adaptation, and fine-grained classification, respectively, on two food datasets.

cs.CV↗

arXiv2025

Towards Unbiased Cross-Modal Representation Learning for Food Image-to-Recipe Retrieval

This paper addresses the challenges of learning representations for recipes and food images in the cross-modal retrieval problem. As the relationship between a recipe and its cooked dish is cause-and-effect, treating a recipe as a text source describing the visual appearance of a dish for learning representation, as the existing approaches, will create bias misleading image-and-recipe similarity judgment. Specifically, a food image may not equally capture every detail in a recipe, due to factors such as the cooking process, dish presentation, and image-capturing conditions. The current representation learning tends to capture dominant visual-text alignment while overlooking subtle variations that determine retrieval relevance. In this paper, we model such bias in cross-modal representation learning using causal theory. The causal view of this problem suggests ingredients as one of the confounder sources and a simple backdoor adjustment can alleviate the bias. By causal intervention, we reformulate the conventional model for food-to-recipe retrieval with an additional term to remove the potential bias in similarity judgment. Based on this theory-informed formulation, we empirically prove the oracle performance of retrieval on the Recipe1M dataset to be MedR=1 across the testing data sizes of 1K, 10K, and even 50K. We also propose a plug-and-play neural module, which is essentially a multi-label ingredient classifier for debiasing. New state-of-the-art search performances are reported on the Recipe1M dataset.

cs.CV↗

arXiv2025

Leveraging Large Language Models for Career Mobility Analysis: A Study of Gender, Race, and Job Change Using U.S. Online Resume Profiles

We present a large-scale analysis of career mobility of college-educated U.S. workers using online resume profiles to investigate how gender, race, and job change options are associated with upward mobility. This study addresses key research questions of how the job changes affect their upward career mobility, and how the outcomes of upward career mobility differ by gender and race. We address data challenges -- such as missing demographic attributes, missing wage data, and noisy occupation labels -- through various data processing and Artificial Intelligence (AI) methods. In particular, we develop a large language models (LLMs) based occupation classification method known as FewSOC that achieves accuracy significantly higher than the original occupation labels in the resume dataset. Analysis of 228,710 career trajectories reveals that intra-firm occupation change has been found to facilitate upward mobility most strongly, followed by inter-firm occupation change and inter-firm lateral move. Women and Black college graduates experience significantly lower returns from job changes than men and White peers. Multilevel sensitivity analyses confirm that these disparities are robust to cluster-level heterogeneity and reveal additional intersectional patterns.

cs.CY↗

arXiv2025

Towards Provably Unlearnable Examples via Bayes Error Optimisation

The recent success of machine learning models, especially large-scale classifiers and language models, relies heavily on training with massive data. These data are often collected from online sources. This raises serious concerns about the protection of user data, as individuals may not have given consent for their data to be used in training. To address this concern, recent studies introduce the concept of unlearnable examples, i.e., data instances that appear natural but are intentionally altered to prevent models from effectively learning from them. While existing methods demonstrate empirical effectiveness, they typically rely on heuristic trials and lack formal guarantees. Besides, when unlearnable examples are mixed with clean data, as is often the case in practice, their unlearnability disappears. In this work, we propose a novel approach to constructing unlearnable examples by systematically maximising the Bayes error, a measurement of irreducible classification error. We develop an optimisation-based approach and provide an efficient solution using projected gradient ascent. Our method provably increases the Bayes error and remains effective when the unlearning examples are mixed with clean samples. Experimental results across multiple datasets and model architectures are consistent with our theoretical analysis and show that our approach can restrict data learnability, effectively in practice.

cs.AI↗

arXiv2025

Mitigating Cross-modal Representation Bias for Multicultural Image-to-Recipe Retrieval

Existing approaches for image-to-recipe retrieval have the implicit assumption that a food image can fully capture the details textually documented in its recipe. However, a food image only reflects the visual outcome of a cooked dish and not the underlying cooking process. Consequently, learning cross-modal representations to bridge the modality gap between images and recipes tends to ignore subtle, recipe-specific details that are not visually apparent but are crucial for recipe retrieval. Specifically, the representations are biased to capture the dominant visual elements, resulting in difficulty in ranking similar recipes with subtle differences in use of ingredients and cooking methods. The bias in representation learning is expected to be more severe when the training data is mixed of images and recipes sourced from different cuisines. This paper proposes a novel causal approach that predicts the culinary elements potentially overlooked in images, while explicitly injecting these elements into cross-modal representation learning to mitigate biases. Experiments are conducted on the standard monolingual Recipe1M dataset and a newly curated multilingual multicultural cuisine dataset. The results indicate that the proposed causal representation learning is capable of uncovering subtle ingredients and cooking actions and achieves impressive retrieval performance on both monolingual and multilingual multicultural datasets.

cs.CV↗

arXiv2025

Seeing Culture: A Benchmark for Visual Reasoning and Grounding

Multimodal vision-language models (VLMs) have made substantial progress in various tasks that require a combined understanding of visual and textual content, particularly in cultural understanding tasks, with the emergence of new cultural datasets. However, these datasets frequently fall short of providing cultural reasoning while underrepresenting many cultures. In this paper, we introduce the Seeing Culture Benchmark (SCB), focusing on cultural reasoning with a novel approach that requires VLMs to reason on culturally rich images in two stages: i) selecting the correct visual option with multiple-choice visual question answering (VQA), and ii) segmenting the relevant cultural artifact as evidence of reasoning. Visual options in the first stage are systematically organized into three types: those originating from the same country, those from different countries, or a mixed group. Notably, all options are derived from a singular category for each type. Progression to the second stage occurs only after a correct visual option is chosen. The SCB benchmark comprises 1,065 images that capture 138 cultural artifacts across five categories from seven Southeast Asia countries, whose diverse cultures are often overlooked, accompanied by 3,178 questions, of which 1,093 are unique and meticulously curated by human annotators. Our evaluation of various VLMs reveals the complexities involved in cross-modal cultural reasoning and highlights the disparity between visual reasoning and spatial grounding in culturally nuanced scenarios. The SCB serves as a crucial benchmark for identifying these shortcomings, thereby guiding future developments in the field of cultural reasoning. https://github.com/buraksatar/SeeingCulture

cs.CV↗

arXiv2025

Effects of Theory of Mind and Prosocial Beliefs on Steering Human-Aligned Behaviors of LLMs in Ultimatum Games

Large Language Models (LLMs) have shown potential in simulating human behaviors and performing theory-of-mind (ToM) reasoning, crucial for complex social interactions. We investigate ToM reasoning's role in aligning agentic behaviors with human norms in negotiation tasks, using the ultimatum game as our referenced task. We initialized LLM agents with different prosocial beliefs (Greedy, Fair, Selfless) and reasoning methods (chain of thought and ToM reasoning of varying levels), examining their decision-making process and outcome across multiple LLMs, including reasoning models like o3-mini and DeepSeek-R1 Distilled Qwen 32B. We perform 2,700 simulations to show that ToM reasoning enhances behavioral alignment with human, decision-making consistency, and negotiation outcomes. Consistent with prior findings, reasoning LLMs exhibit limited capability compared to ToM-enhanced LLMs, with different game roles benefiting from different ToM orders. Fair proposers and responders accepting offers were the most consistent with their strategic reasonings, whereas all agents showed strong consistencies with human beliefs when rejecting offers, except when the offer was fair. Human verification further revealed that Llama 3.3 70B produces reasoning most consistent with its actions and beliefs. Our findings advance understanding of ToM's role in human-AI interaction and cooperative decision-making. The code used for our experiments can be found at https://github.com/Stealth-py/UltimatumToM.

cs.CL↗

arXiv2025

Advancing Food Nutrition Estimation via Visual-Ingredient Feature Fusion

Nutrition estimation is an important component of promoting healthy eating and mitigating diet-related health risks. Despite advances in tasks such as food classification and ingredient recognition, progress in nutrition estimation is limited due to the lack of datasets with nutritional annotations. To address this issue, we introduce FastFood, a dataset with 84,446 images across 908 fast food categories, featuring ingredient and nutritional annotations. In addition, we propose a new model-agnostic Visual-Ingredient Feature Fusion (VIF$^2$) method to enhance nutrition estimation by integrating visual and ingredient features. Ingredient robustness is improved through synonym replacement and resampling strategies during training. The ingredient-aware visual feature fusion module combines ingredient features and visual representation to achieve accurate nutritional prediction. During testing, ingredient predictions are refined using large multimodal models by data augmentation and majority voting. Our experiments on both FastFood and Nutrition5k datasets validate the effectiveness of our proposed method built in different backbones (e.g., Resnet, InceptionV3 and ViT), which demonstrates the importance of ingredient information in nutrition estimation. https://huiyanqi.github.io/fastfood-nutrition-estimation/.

cs.CV↗

arXiv2025

A Multi-Stage Framework with Taxonomy-Guided Reasoning for Occupation Classification Using Large Language Models

Automatically annotating job data with standardized occupations from taxonomies, known as occupation classification, is crucial for labor market analysis. However, this task is often hindered by data scarcity and the challenges of manual annotations. While large language models (LLMs) hold promise due to their extensive world knowledge and in-context learning capabilities, their effectiveness depends on their knowledge of occupational taxonomies, which remains unclear. In this study, we assess the ability of LLMs to generate precise taxonomic entities from taxonomy, highlighting their limitations, especially for smaller models. To address these challenges, we propose a multi-stage framework consisting of inference, retrieval, and reranking stages, which integrates taxonomy-guided reasoning examples to enhance performance by aligning outputs with taxonomic knowledge. Evaluations on a large-scale dataset show that our framework not only enhances occupation and skill classification tasks, but also provides a cost-effective alternative to frontier models like GPT-4o, significantly reducing computational costs while maintaining strong performance. This makes it a practical and scalable solution for occupation classification and related tasks across LLMs.

cs.CL↗

arXiv2025

Benchmarking Gaslighting Negation Attacks Against Multimodal Large Language Models

Multimodal Large Language Models (MLLMs) have exhibited remarkable advancements in integrating different modalities, excelling in complex understanding and generation tasks. Despite their success, MLLMs remain vulnerable to conversational adversarial inputs. In this paper, we systematically study gaslighting negation attacks: a phenomenon where models, despite initially providing correct answers, are persuaded by user-provided negations to reverse their outputs, often fabricating justifications. We conduct extensive evaluations of state-of-the-art MLLMs across diverse benchmarks and observe substantial performance drops when negation is introduced. Notably, we introduce the first benchmark GaslightingBench, specifically designed to evaluate the vulnerability of MLLMs to negation arguments. GaslightingBench consists of multiple-choice questions curated from existing datasets, along with generated negation prompts across 20 diverse categories. Throughout extensive evaluation, we find that proprietary models such as Gemini-1.5-flash and GPT-4o demonstrate better resilience compared to open-source counterparts like Qwen2-VL and LLaVA, though even advanced reasoning-oriented models like Gemini-2.5-Pro remain susceptible. Our category-level analysis further shows that subjective or socially nuanced domains (e.g., Social Relation, Image Emotion) are especially fragile, while more objective domains (e.g., Geography) exhibit relatively smaller but still notable drops. Overall, all evaluated MLLMs struggle to maintain logical consistency under gaslighting negation attack. These findings highlight a fundamental robustness gap and provide insights for developing more reliable and trustworthy multimodal AI systems. Project website: https://yxg1005.github.io/GaslightingNegationAttacks/.

cs.CL↗

arXiv2024

MathHay: An Automated Benchmark for Long-Context Mathematical Reasoning in LLMs

Recent large language models (LLMs) have demonstrated versatile capabilities in long-context scenarios. Although some recent benchmarks have been developed to evaluate the long-context capabilities of LLMs, there is a lack of benchmarks evaluating the mathematical reasoning abilities of LLMs over long contexts, which is crucial for LLMs' application in real-world scenarios. In this paper, we introduce MathHay, an automated benchmark designed to assess the long-context mathematical reasoning capabilities of LLMs. Unlike previous benchmarks like Needle in a Haystack, which focus primarily on information retrieval within long texts, MathHay demands models with both information-seeking and complex mathematical reasoning abilities. We conduct extensive experiments on MathHay to assess the long-context mathematical reasoning abilities of eight top-performing LLMs. Even the best-performing model, Gemini-1.5-Pro-002, still struggles with mathematical reasoning over long contexts, achieving only 51.26% accuracy at 128K tokens. This highlights the significant room for improvement on the MathHay benchmark.

cs.CL↗

arXiv2024

Speaker Verification in Agent-Generated Conversations

The recent success of large language models (LLMs) has attracted widespread interest to develop role-playing conversational agents personalized to the characteristics and styles of different speakers to enhance their abilities to perform both general and special purpose dialogue tasks. However, the ability to personalize the generated utterances to speakers, whether conducted by human or LLM, has not been well studied. To bridge this gap, our study introduces a novel evaluation challenge: speaker verification in agent-generated conversations, which aimed to verify whether two sets of utterances originate from the same speaker. To this end, we assemble a large dataset collection encompassing thousands of speakers and their utterances. We also develop and evaluate speaker verification models under experiment setups. We further utilize the speaker verification models to evaluate the personalization abilities of LLM-based role-playing models. Comprehensive experiments suggest that the current role-playing models fail in accurately mimicking speakers, primarily due to their inherent linguistic characteristics.

cs.CL↗