As scaling up training data has significantly improved the general multimodal capabilities of Large Vision-Language Models (LVLMs), they still suffer from the hallucination issue, generating text that is inconsistent with the visual input. This phenomenon motivates us to systematically investigate the role of training data in hallucination. We introduce a new benchmark, POPEv2, which consists of counterfactual images collected from the training data of LVLMs with certain objects masked. Through comprehensive evaluation on POPEv2, we find that current LVLMs suffer from training bias: they fail to fully leverage their training data and hallucinate more frequently on images seen during training. Specifically, they perform poorly on counterfactual images, often incorrectly answering ``Yes'' to questions about masked objects. To understand this issue, we conduct probing experiments on the models' internal components, revealing that this training bias is primarily located in the language modeling (LM) head. Based on these findings, we propose Obliviate, an efficient and lightweight unlearning method designed to mitigate object hallucination via training bias unlearning. Obliviate identifies the discrepancy between ground-truth labels and model outputs on the training data as a proxy for bias and adopts a parameter- and data-efficient fine-tuning strategy that only updates the LM head. Extensive experiments demonstrate the effectiveness of our approach. While only reusing the training data and updating approximately 2\% of the parameters, Obliviate significantly reduces hallucination across both discriminative and generative tasks. Furthermore, it demonstrates strong scalability with respect to both model size (2B to 72B) and training data volume, and exhibits promising generalization to hallucination types beyond object-level hallucination. Our code and data will be publicly released.

Fostering students' abilities for knowledge integration and transfer in complex problem-solving scenarios is a core objective of modern education, and interdisciplinary STEM is a key pathway to achieve this, yet it requires expert guidance that is difficult to scale. While LLMs offer potential in this regard, their true capability for guided instruction remains unclear due to the lack of an effective evaluation benchmark. To address this, we introduce SID, the first benchmark designed to systematically evaluate the higher-order guidance capabilities of LLMs in multi-turn, interdisciplinary Socratic dialogues. Our contributions include a large-scale dataset of 10,000 dialogue turns across 48 complex STEM projects, a novel annotation schema for capturing deep pedagogical features, and a new suite of evaluation metrics (e.g., X-SRG). Baseline experiments confirm that even state-of-the-art LLMs struggle to execute effective guided dialogues that lead students to achieve knowledge integration and transfer. This highlights the critical value of our benchmark in driving the development of more pedagogically-aware LLMs.

The rapid advancement of AI-generation models has enabled the creation of hyperrealistic imagery, posing ethical risks through widespread misinformation. Current deepfake detection methods, categorized as face specific detectors or general AI-generated detectors, lack transparency by framing detection as a classification task without explaining decisions. While several LLM-based approaches offer explainability, they suffer from coarse-grained analyses and dependency on labor-intensive annotations. This paper introduces RAIDX (Retrieval-Augmented Image Deepfake Detection and Explainability), a novel deepfake detection framework integrating Retrieval-Augmented Generation (RAG) and Group Relative Policy Optimization (GRPO) to enhance detection accuracy and decision explainability. Specifically, RAIDX leverages RAG to incorporate external knowledge for improved detection accuracy and employs GRPO to autonomously generate fine-grained textual explanations and saliency maps, eliminating the need for extensive manual annotations. Experiments on multiple benchmarks demonstrate RAIDX's effectiveness in identifying real or fake, and providing interpretable rationales in both textual descriptions and saliency maps, achieving state-of-the-art detection performance while advancing transparency in deepfake identification. RAIDX represents the first unified framework to synergize RAG and GRPO, addressing critical gaps in accuracy and explainability. Our code and models will be publicly available.

The dream to create AI assistants as capable and versatile as the fictional J.A.R.V.I.S from Iron Man has long captivated imaginations. With the evolution of (multi-modal) large language models ((M)LLMs), this dream is closer to reality, as (M)LLM-based Agents using computing devices (e.g., computers and mobile phones) by operating within the environments and interfaces (e.g., Graphical User Interface (GUI)) provided by operating systems (OS) to automate tasks have significantly advanced. This paper presents a comprehensive survey of these advanced agents, designated as OS Agents. We begin by elucidating the fundamentals of OS Agents, exploring their key components including the environment, observation space, and action space, and outlining essential capabilities such as understanding, planning, and grounding. We then examine methodologies for constructing OS Agents, focusing on domain-specific foundation models and agent frameworks. A detailed review of evaluation protocols and benchmarks highlights how OS Agents are assessed across diverse tasks. Finally, we discuss current challenges and identify promising directions for future research, including safety and privacy, personalization and self-evolution. This survey aims to consolidate the state of OS Agents research, providing insights to guide both academic inquiry and industrial development. An open-source GitHub repository is maintained as a dynamic resource to foster further innovation in this field. We present a 9-page version of our work, accepted by ACL 2025, to provide a concise overview to the domain.

Red teaming is critical for identifying vulnerabilities and building trust in current LLMs. However, current automated methods for Large Language Models (LLMs) rely on brittle prompt templates or single-turn attacks, failing to capture the complex, interactive nature of real-world adversarial dialogues. We propose a novel paradigm: training an AI to strategically `break' another AI. By formalizing red teaming as a Markov Decision Process (MDP) and employing a hierarchical Reinforcement Learning (RL) framework, we effectively address the inherent sparse reward and long-horizon challenges. Our generative agent learns coherent, multi-turn attack strategies through a fine-grained, token-level harm reward, enabling it to uncover subtle vulnerabilities missed by existing baselines. This approach sets a new state-of-the-art, fundamentally reframing LLM red teaming as a dynamic, trajectory-based process (rather than a one-step test) essential for robust AI deployment.

This paper addresses the critical need for scalable and high-quality educational assessment tools within the Malaysian education system. It highlights the potential of Generative AI (GenAI) while acknowledging the significant challenges of ensuring factual accuracy and curriculum alignment, especially for low-resource languages like Bahasa Melayu. This research introduces and compares four incremental pipelines for generating Form 1 Mathematics multiple-choice questions (MCQs) in Bahasa Melayu using OpenAI's GPT-4o. The methods range from non-grounded prompting (structured and basic) to Retrieval-Augmented Generation (RAG) approaches (one using the LangChain framework, one implemented manually). The system is grounded in official curriculum documents, including teacher-prepared notes and the yearly teaching plan (RPT). A dual-pronged automated evaluation framework is employed to assess the generated questions. Curriculum alignment is measured using Semantic Textual Similarity (STS) against the RPT, while contextual validity is verified through a novel RAG-based Question-Answering (RAG-QA) method. The results demonstrate that RAG-based pipelines significantly outperform non-grounded prompting methods, producing questions with higher curriculum alignment and factual validity. The study further analyzes the trade-offs between the ease of implementation of framework-based RAG and the fine-grained control offered by a manual pipeline. This work presents a validated methodology for generating curriculum-specific educational content in a low-resource language, introduces a symbiotic RAG-QA evaluation technique, and provides actionable insights for the development and deployment of practical EdTech solutions in Malaysia and similar regions.

High-quality, multi-turn instructional dialogues between novices and experts are essential for developing AI systems that support teaching, learning, and decision-making. These dialogues often involve scaffolding -- the process by which an expert supports a novice's thinking through questions, feedback, and step-by-step guidance. However, such data are scarce due to privacy concerns in recording and the vulnerability inherent in help-seeking. We present SimInstruct, a scalable, expert-in-the-loop tool for collecting scaffolding dialogues. Using teaching development coaching as an example domain, SimInstruct simulates novice instructors via LLMs, varying their teaching challenges and LLM's persona traits, while human experts provide multi-turn feedback, reasoning, and instructional support. This design enables the creation of realistic, pedagogically rich dialogues without requiring real novice participants. Our results reveal that persona traits, such as extroversion and introversion, meaningfully influence how experts engage. Compared to real mentoring recordings, SimInstruct dialogues demonstrate comparable pedagogical relevance and cognitive depth. Experts also reported the process as engaging and reflective, improving both data quality and their own professional insight. We further fine-tuned a LLaMA model to be an expert model using the augmented dataset, which outperformed GPT-4o in instructional quality. Our analysis highlights GPT-4o's limitations in weak reflective questioning, overuse of generic praise, a condescending tone, and a tendency to overwhelm novices with excessive suggestions.

Multimodal learning has witnessed remarkable advancements in recent years, particularly with the integration of attention-based models, leading to significant performance gains across a variety of tasks. Parallel to this progress, the demand for explainable artificial intelligence (XAI) has spurred a growing body of research aimed at interpreting the complex decision-making processes of these models. This systematic literature review analyzes research published between January 2020 and early 2024 that focuses on the explainability of multimodal models. Framed within the broader goals of XAI, we examine the literature across multiple dimensions, including model architecture, modalities involved, explanation algorithms and evaluation methodologies. Our analysis reveals that the majority of studies are concentrated on vision-language and language-only models, with attention-based techniques being the most commonly employed for explanation. However, these methods often fall short in capturing the full spectrum of interactions between modalities, a challenge further compounded by the architectural heterogeneity across domains. Importantly, we find that evaluation methods for XAI in multimodal settings are largely non-systematic, lacking consistency, robustness, and consideration for modality-specific cognitive and contextual factors. Based on these findings, we provide a comprehensive set of recommendations aimed at promoting rigorous, transparent, and standardized evaluation and reporting practices in multimodal XAI research. Our goal is to support future research in more interpretable, accountable, and responsible mulitmodal AI systems, with explainability at their core.

Referring Audio-Visual Segmentation (Ref-AVS) aims to segment target objects in audible videos based on given reference expressions. Prior works typically rely on learning latent embeddings via multimodal fusion to prompt a tunable SAM/SAM2 decoder for segmentation, which requires strong pixel-level supervision and lacks interpretability. From a novel perspective of explicit reference understanding, we propose TGS-Agent, which decomposes the task into a Think-Ground-Segment process, mimicking the human reasoning procedure by first identifying the referred object through multimodal analysis, followed by coarse-grained grounding and precise segmentation. To this end, we first propose Ref-Thinker, a multimodal language model capable of reasoning over textual, visual, and auditory cues. We construct an instruction-tuning dataset with explicit object-aware think-answer chains for Ref-Thinker fine-tuning. The object description inferred by Ref-Thinker is used as an explicit prompt for Grounding-DINO and SAM2, which perform grounding and segmentation without relying on pixel-level supervision. Additionally, we introduce R\textsuperscript{2}-AVSBench, a new benchmark with linguistically diverse and reasoning-intensive references for better evaluating model generalization. Our approach achieves state-of-the-art results on both standard Ref-AVSBench and proposed R\textsuperscript{2}-AVSBench. Code will be available at https://github.com/jasongief/TGS-Agent.

The remarkable success of Large Language Models (LLMs) in generative tasks has raised fundamental questions about the nature of their acquired capabilities, which often appear to emerge unexpectedly without explicit training. This paper examines the emergent properties of Deep Neural Networks (DNNs) through both theoretical analysis and empirical observation, addressing the epistemological challenge of "creation without understanding" that characterises contemporary AI development. We explore how the neural approach's reliance on nonlinear, stochastic processes fundamentally differs from symbolic computational paradigms, creating systems whose macro-level behaviours cannot be analytically derived from micro-level neuron activities. Through analysis of scaling laws, grokking phenomena, and phase transitions in model capabilities, I demonstrate that emergent abilities arise from the complex dynamics of highly sensitive nonlinear systems rather than simply from parameter scaling alone. My investigation reveals that current debates over metrics, pre-training loss thresholds, and in-context learning miss the fundamental ontological nature of emergence in DNNs. I argue that these systems exhibit genuine emergent properties analogous to those found in other complex natural phenomena, where systemic capabilities emerge from cooperative interactions among simple components without being reducible to their individual behaviours. The paper concludes that understanding LLM capabilities requires recognising DNNs as a new domain of complex dynamical systems governed by universal principles of emergence, similar to those operating in physics, chemistry, and biology. This perspective shifts the focus from purely phenomenological definitions of emergence to understanding the internal dynamic transformations that enable these systems to acquire capabilities that transcend their individual components.