Showing 1–50 of 162 results for author: Lim, B

AlignFace: Human-Aligned Face Similarity Metric with Interpretable Concept Relations

Authors: Ying Huang, Wencan Zhang, Brian Y. Lim

Abstract: Computer vision models for generated facial content, such as face editing and privacy protection, increasingly affect people, requiring similarity metrics that serve as faithful proxies for human perception. While perceptual evaluation has progressed from signal-based heuristics to representation-based metrics, current approaches are limited to behavioral modeling without cognitive alignment. They… ▽ More Computer vision models for generated facial content, such as face editing and privacy protection, increasingly affect people, requiring similarity metrics that serve as faithful proxies for human perception. While perceptual evaluation has progressed from signal-based heuristics to representation-based metrics, current approaches are limited to behavioral modeling without cognitive alignment. They rely on implicit and spurious relations while assuming a universal observer, failing to account for inherent variations across diverse human populations. This leads to inaccurate evaluative models of stakeholders and misleading guidance for generative model debugging. Rather than treating perception as a black box, we leverage scientific findings from cognitive psychology of human face similarity perception: dependence on facial featural and configural attributes, nonlinear psychophysical response scaling, and own-group biases. We introduce the FACETS dataset and propose AlignFace, an interpretable, human-aligned, face similarity metric that encodes these cognitive principles through ante-hoc modeling. It employs visual-language modeling (VLM) to encode paired face images and text-based attributes, gated cross-attention (CA) to extract attribute-specific facial difference representations, concept bottleneck modeling (CBM) to constrain reasoning via interpretable face attributes, and neural generalized additive model (GAM) to model their nonlinear influence. Experiments found AlignFace significantly improves alignment with human subpopulation perceptions compared to baseline metrics, including recent domain-free learned perceptual metrics. By bridging learned representations and human cognitive processes, this work enables more transparent and aligned perceptual evaluation metrics for face images. △ Less

Submitted 14 August, 2026; originally announced August 2026.

Comments: 10 pages, 10 figures, 2 tables, ACM MM 26

Persuasive and Compliant Tendencies Predict Group Decision-Making in Humans and Language Models

Authors: Wenwen He, Wenke Huang, Wei Yang Bryan Lim, Dacheng Tao

Abstract: Large language models (LLMs) are increasingly involved in group decision-making with other LLMs and humans. Yet it remains unclear whether their influence is driven by persuasion-oriented expression or compliance-oriented accommodation. We introduce DecisionQE, a questionnaire-based framework for measuring each model's persuasive and compliant tendencies across multiple decision scenarios, and use… ▽ More Large language models (LLMs) are increasingly involved in group decision-making with other LLMs and humans. Yet it remains unclear whether their influence is driven by persuasion-oriented expression or compliance-oriented accommodation. We introduce DecisionQE, a questionnaire-based framework for measuring each model's persuasive and compliant tendencies across multiple decision scenarios, and use the Werewolf game as an interactive testbed to study their effects on social influence and group outcomes under asymmetric information. Across experiments, stronger persuasive tendency does not significantly improve group outcomes, whereas compliant-oriented models show more stable advantages in cooperation. We further reveal a dual effect of compliance: it supports cooperation in honest roles but improves concealment in adversarial roles. These findings suggest that LLM group interactions reveal not only task outcomes, but also measurable patterns of intrinsic behavioral tendency. LLMs can therefore serve as a lens for sociological observation of language-mediated interaction, while highlighting the need to incorporate behavioral tendencies into safety evaluation of LLM systems. △ Less

Submitted 8 August, 2026; originally announced August 2026.

SynChain: Inducing Computer-Use Agent Systems to Construct Their Own Attack Chains

Authors: Fuyao Zhang, Jiaming Zhang, Che Wang, Boyang Chen, Yurong Hao, Xiongtao Sun, Guowei Guan, Blaise Delattre, Yang Cao, Wei Yang Bryan Lim

Abstract: Computer-use agents~(CUAs) have transformed large language models into persistent execution systems capable of generating, storing, and reusing artifacts like skills and memory entries. However, existing security defenses largely treat attacks as externally triggered or temporally bounded, leaving a critical gap in addressing how compromise can propagate internally through an agent's own persisten… ▽ More Computer-use agents~(CUAs) have transformed large language models into persistent execution systems capable of generating, storing, and reusing artifacts like skills and memory entries. However, existing security defenses largely treat attacks as externally triggered or temporally bounded, leaving a critical gap in addressing how compromise can propagate internally through an agent's own persistent state. We reveal that malicious influence can be covertly embedded into the structural redundancies of autonomously synthesized artifacts, allowing it to survive internal state updates and bypass standard vetting mechanisms. To formalize this threat, we introduce SynChain, a self-synthesized attack paradigm utilizing persistence-aware directed supervised fine-tuning to induce agents to create poisoned yet benign-looking artifacts. To systematically evaluate this propagation, we construct CUAChain, a dataset comprising 30 benign task chains and three attack objectives. SynChain enables dormant payloads to seamlessly reactivate in future workflows as trusted context, operating entirely without new malicious exogenous inputs. Extensive experiments on OpenClaw, Codex, and Claude Code under four defense settings demonstrate that SynChain achieves high attack success and outperforms adapted baselines, proving that securing CUAs requires provenance-aware reasoning over cross-task execution trajectories. △ Less

Submitted 7 August, 2026; originally announced August 2026.

Adversarial Attacks for Good: A Survey of Proactive Protection across the Visual Content Lifecycle

Authors: Jiaming Zhang, Boyang Chen, Zherui Li, Fuyao Zhang, Xinyu Yan, Hong Xi Tae, Wenwen He, Xuan Wang, Siqi Guo, Junhao Dong, Kun Wang, Hanxun Huang, Yige Li, Xingjun Ma, Yang Cao, Lingjuan Lyu, Wei Yang Bryan Lim

Abstract: Once visual content enters an AI pipeline, its owner often retains little technical control over how it is used. Legal and regulatory remedies can address misuse, but many technical interventions must be applied earlier, when content is released or accessed. This survey examines the protective paradigm that has grown around this intervention point, which we call \emph{adversarial attacks for good}… ▽ More Once visual content enters an AI pipeline, its owner often retains little technical control over how it is used. Legal and regulatory remedies can address misuse, but many technical interventions must be applied earlier, when content is released or accessed. This survey examines the protective paradigm that has grown around this intervention point, which we call \emph{adversarial attacks for good}. Perturbations and structured signals long studied as attacks on learned models are instead applied by data owners, creators, platforms, or auditors to disrupt unauthorized automation or support later accountability. Five research communities have arrived at this inversion largely independently, each addressing a different stage of a visual asset's lifecycle: privacy filters against unwanted recognition at sharing time, unlearnable examples against unauthorized training, generative safeguards against malicious editing or imitation, adversarial CAPTCHAs for access control against automated agents, and provenance mechanisms for post-circulation attribution. Although developed in separate venues with incompatible success criteria, many of these methods exploit persistent gaps between human perception, semantic interpretation, and machine inference, suggesting that the paradigm remains relevant as visual pipelines evolve toward multimodal models and autonomous agents. To make their claims comparable, we evaluate all five families along shared axes of transferability, adaptability, and deployment readiness. Across the lifecycle, we find that most protections are still validated mainly against static or weakly adaptive adversaries, while evidence beyond controlled benchmarks remains scarce. We close by consolidating cross-stage countermeasures and open problems for robust, composable, and deployable owner-side protection. △ Less

Submitted 4 August, 2026; originally announced August 2026.

CoCurve: Cross-Module Co-Pruning Curvature for Training-Free Structured LLM Pruning

Authors: Zhiren Gong, Zihao Zeng, Zijie Wang, Tiantong Wang, Chau Yuen, Wei Yang Bryan Lim

Abstract: Structured pruning compresses large language models (LLMs) by removing whole computational units, such as attention heads and feed-forward (FFN) channel groups. Most training-free methods, however, rank these units independently, implicitly treating the loss from pruning a set as the sum of its individual losses. This view fails for Transformers, whose sublayers are coupled through a shared residu… ▽ More Structured pruning compresses large language models (LLMs) by removing whole computational units, such as attention heads and feed-forward (FFN) channel groups. Most training-free methods, however, rank these units independently, implicitly treating the loss from pruning a set as the sum of its individual losses. This view fails for Transformers, whose sublayers are coupled through a shared residual stream. Two individually weak units can thus be jointly indispensable, yet independent scoring is blind to such dependence and removes them together. We introduce CoCurve (Cross-Module Co-Pruning Curvature), a calibration-only, fine-tuning-free method that prunes attention and FFN units jointly. A second-order Taylor expansion of the token-level KL between the frozen model and its masked copy yields a single Fisher matrix whose diagonal is classical node saliency and whose off-diagonal entries are co-pruning curvature edges: the extra damage of removing two units together. Under a single-ablation additivity approximation this matrix reduces to a Gram product of single-unit ablation features, so the full M x M interaction is recovered from M forward passes, with no pairwise sweeps or gradients. Pruning then reduces to one budgeted quadratic program, solved in a single shot under a shared attention--FFN budget, with no labels, fine-tuning, or recovery. △ Less

Submitted 20 July, 2026; originally announced July 2026.

YeTI: You Only Need Two Noisy Images for Real-World sRGB Noise Generation

Authors: Jaekyun Ko, Byung Wan Lim, Soomin Lee, Dongjin Kim, Tae Hyun Kim

Abstract: Real-world sRGB image denoising remains challenging due to the nonlinear characteristics of sensor noise and the difficulty of acquiring aligned clean-noisy image pairs. Supervised denoisers often overfit to limited paired datasets, while self-supervised methods still depend on sufficiently diverse noisy observations. These limitations motivate scalable noise synthesis methods that can model real-… ▽ More Real-world sRGB image denoising remains challenging due to the nonlinear characteristics of sensor noise and the difficulty of acquiring aligned clean-noisy image pairs. Supervised denoisers often overfit to limited paired datasets, while self-supervised methods still depend on sufficiently diverse noisy observations. These limitations motivate scalable noise synthesis methods that can model real-world noise without clean ground truth or camera metadata. We propose YeTI, a real-world sRGB noise generation framework that learns from only two noisy observations of the same scene. YeTI uses a Reconstruction Autoencoder to disentangle scene structure and noise characteristics, and models the latent noise distribution with a one-step Conditional Diffusion Transformer trained using consistency objectives. Given a single noisy input at inference time, YeTI generates realistic, signal-dependent noise while preserving the underlying scene content. Extensive experiments demonstrate the effectiveness of YeTI across real-world benchmarks. We evaluate noise generation on SIDD and further assess generalization on SIDD+, MAI2021, and SID, covering smartphone and diverse consumer-camera sensors. Downstream denoising results on DND further show that denoisers trained with YeTI-synthesized images achieve strong real-world performance, highlighting the practical value of clean-image-free and metadata-free noise generation. △ Less

Submitted 10 July, 2026; originally announced July 2026.

Comments: Accepted to ECCV 2026. Includes supplementary material

Conditional Co-Ablation: Recovering Self-Repair Backups in Transformer Circuits

Authors: Zhiren Gong, Zihao Zeng, Chau Yuen, Wei Yang Bryan Lim

Abstract: Mechanistic interpretability often relies on component-level interventions to discover how a model produces a behavior. This guides attribution, capability knockout, and model pruning downstream to operate by scoring each unit by the effect of ablation in isolation. Such first-order scoring is natural when component importance is additive, but becomes misleading when a transformer self-repairs: af… ▽ More Mechanistic interpretability often relies on component-level interventions to discover how a model produces a behavior. This guides attribution, capability knockout, and model pruning downstream to operate by scoring each unit by the effect of ablation in isolation. Such first-order scoring is natural when component importance is additive, but becomes misleading when a transformer self-repairs: after a primary component is removed, a dormant backup can take over, muting the primary's measured effect while the backup itself appears irrelevant on the intact model. We recast this failure as a recovery task, conditional circuit completion, and introduce Conditional Co-Ablation (CoAx), a label-free, output-grounded score that asks how much each remaining unit's ablation effect grows once a primary set has been removed. This conditional growth exposes the second-order interaction that single-unit scores discard. On the GPT-2-small IOI circuit, CoAx raises backup-head recovery from 0.33 to 0.91 ROC-AUC, outperforming all baselines, including self-repair-aware gradient scores (best 0.82); counterfactual patching verifies that the recovered heads causally carry the repair. The same label-free procedure transfers to induction across eight models. Beyond discovery, the recovered backups correct self-repair-masked attribution, identify the components required for capability knockout, and yield repair-aware structured pruning scaling from 124M to 7B. Component importance is therefore not merely an isolated-unit property: in robust circuits, the components that matter can become visible only under the interventions that make them necessary. △ Less

Submitted 2 July, 2026; originally announced July 2026.

SketchXplain: Intuitive Visual Explanations of Image Classifiers with Sketches

Authors: Wencan Zhang, Mario Michelessa, Xuejun Zhao, Brian Y. Lim

Abstract: Saliency map visualizations explain image-based AI predictions by pointing to regions, but these are often unintuitive and semantically unclear, leaving an interpretability gap. We argue that AI explanations should be intuitive -- coherent to user knowledge, yet simple and selective to accelerate interpretation. Inspired by artistic drawings, we propose SketchXplain to generate sketch-based visual… ▽ More Saliency map visualizations explain image-based AI predictions by pointing to regions, but these are often unintuitive and semantically unclear, leaving an interpretability gap. We argue that AI explanations should be intuitive -- coherent to user knowledge, yet simple and selective to accelerate interpretation. Inspired by artistic drawings, we propose SketchXplain to generate sketch-based visual explanations for intuitive image-based explainable AI (XAI). Combining techniques in saliency maps, concept-bottleneck models, and sketch optimization, SketchXplain integrates saliency to select coherent observation artifacts, concepts for knowledge coherence, cues to represent them, and abstraction for simplicity. Evaluating on face expression recognition, modeling and user studies showed that SketchXplain supported quicker interpretation with more aligned visualizations than saliency maps or simple drawings. Further evaluation on skin lesion diagnosis found that SketchXplain more coherently visualized disease symptoms, better supporting lay diagnosis. Thus, this work illustrates the value of sketches for intuitive, simple, coherent, and quick image-based XAI visualizations. △ Less

Submitted 16 June, 2026; originally announced June 2026.

Comments: 14 pages, 6 figures, 4 tables. Submitted to TVCG

ACM Class: H.5.2; I.2

Keeping an Eye on AI: A Framework for Effective Human Oversight of AI Systems

Authors: Susanne Gaube, Markus Langer, Tim Miller, Kevin Baum, Raimund Dachselt, Anna Maria Feit, Ujwal Gadiraju, Harmanpreet Kaur, Mark T. Keane, Richard Landers, Johann Laux, Q. Vera Liao, Brian Lim, Linda Onnasch, Tim Schrills, Liz Sonenberg, Chenhao Tan, Nava Tintarev, Ziang Xiao, Hanwei Zhang

Abstract: The use of Artificial Intelligence (AI) in high-risk, decision-making scenarios presents technical, safety, and normative challenges; problems that may only be ameliorated by human oversight. However, notions of human oversight lack a common foundational understanding: oversight architectures are not well defined, the roles involved remain unclear, and implementation steps are opaque. Hence, resea… ▽ More The use of Artificial Intelligence (AI) in high-risk, decision-making scenarios presents technical, safety, and normative challenges; problems that may only be ameliorated by human oversight. However, notions of human oversight lack a common foundational understanding: oversight architectures are not well defined, the roles involved remain unclear, and implementation steps are opaque. Hence, researchers and practitioners struggle to determine how to design, implement, and evaluate systems that enable effective human oversight. This paper advances a practical framework for effective human oversight of AI systems, based on a cross-disciplinary perspective that draws on insights from computer science, human-computer interaction, psychology, philosophy, and law. The core contributions are: (1) a foundational framework, with a working definition, architecture and processes for effective human oversight of AI systems; (2) an initial template for documenting oversight architectures and processes, applied to diverse domains; and (3) a synthesis of open research challenges that need to be considered in the emerging field of effective human oversight of AI systems. △ Less

Submitted 9 April, 2026; originally announced May 2026.

Comments: The conceptual analysis for this work was undertaken by the authors at Dagstuhl seminar 25272 'Challenges of Human Oversight: Achieving Human Control of AI-Based Systems' (https://www.dagstuhl.de/25272), held at Schloss Dagstuhl (June 29th-July 4th, 2025)

XDomainBench: Diagnosing Reasoning Collapse in High-Dimensional Scientific Knowledge Composition

Authors: Gong Zhiren, Tiantong Wu, Jiaming Zhang, Fuyao Zhang, Che Wang, Yurong Hao, Yikun Hou, Foo Ping, Yilei Zhao, Fei Huang, Chau Yuen, Wei Yang Bryan Lim

Abstract: Large Language Models (LLMs) are increasingly deployed for knowledge synthesis, yet their capacity for compositional generalization in scientific knowledge remains under-characterized. Existing benchmarks primarily focus on single-turn restricted scenarios, failing to capture the capability boundaries exposed by real-world interactive scientific workflows. To address this, we introduce XDomainBenc… ▽ More Large Language Models (LLMs) are increasingly deployed for knowledge synthesis, yet their capacity for compositional generalization in scientific knowledge remains under-characterized. Existing benchmarks primarily focus on single-turn restricted scenarios, failing to capture the capability boundaries exposed by real-world interactive scientific workflows. To address this, we introduce XDomainBench, a diagnostic benchmark for interactive interdisciplinary scientific reasoning. We formalize the composition order and mixture structure to enable systematic stress-testing from single-discipline to inter-disciplinary, comprising 8,598 interactive sessions across 20 domains and 4 task categories, with 8 realistic trajectory patterns covering difficulty and domain-mixture dynamics, simulating real AI4S scenarios. Large-scale evaluation of LLMs reveals a systematic reasoning collapse as composition order increases, stemming from two root causes: (i) direct difficulty increases induced by domain composition, and (ii) indirect interaction-amplified failures where trajectory patterns trigger error accumulation, reasoning breaks, and domain confusion, ultimately leading to session collapse. △ Less

Submitted 14 May, 2026; originally announced May 2026.

SR-Platform: An Agentic Pipeline for Natural Language-Driven Robot Simulation Environment Synthesis

Authors: Ben Wei Lim, Minh Duc Le, Thang Truong, Thanh Nguyen Canh

Abstract: Generating robot simulation environments remains a major bottleneck in simulation-based robot learning. Constructing a training-ready MuJoCo scene typically requires expertise in 3D asset modeling, MJCF specification, spatial layout, collision avoidance, and robot-model integration. We present SR-Platform, a production-deployed agentic system that converts free-form natural language descriptions i… ▽ More Generating robot simulation environments remains a major bottleneck in simulation-based robot learning. Constructing a training-ready MuJoCo scene typically requires expertise in 3D asset modeling, MJCF specification, spatial layout, collision avoidance, and robot-model integration. We present SR-Platform, a production-deployed agentic system that converts free-form natural language descriptions into executable, physically valid MuJoCo environments. SR-Platform decomposes scene synthesis into four stages: an LLM-based orchestrator that converts user intent into a structured scene plan; an asset forge that retrieves cached assets or generates new 3D geometry through LLM-to-CadQuery synthesis; a layout architect that assigns object poses and verifies industrial constraints; and a bridge layer that assembles the final MJCF scene and merges the selected robot model. The system is deployed as a nine-service Docker stack with WebSocket progress streaming, MinIO-backed mesh storage, Qdrant-based semantic asset retrieval, Redis job state, and InfluxDB telemetry. Using 30 days of production telemetry covering 611 successful LLM calls, SR-Platform generates five-object scenes with a median end-to-end latency of approximately 50 s, while cache-accelerated scenes complete in approximately 30-40 s. The asset forge shows an 11.3% first-attempt retry rate with automatic recovery, and cached asset retrieval removes per-object LLM calls for previously generated object types. These results show that agentic scene synthesis can reduce the manual effort required to create diverse robot training environments, enabling users to produce executable MuJoCo scenes from plain English prompts in under one minute. △ Less

Submitted 14 May, 2026; originally announced May 2026.

Certified Robustness under Heterogeneous Perturbations via Hybrid Randomized Smoothing

Authors: Blaise Delattre, Hengyu Wu, Paul Caillon, Wei Yang Bryan Lim, Yang Cao

Abstract: Randomized smoothing provides strong, model-agnostic robustness certificates, but existing guarantees are limited to single modalities, treating continuous and discrete inputs in isolation. This limitation becomes critical in multimodal models, where decisions depend on cross-modal semantics and adversaries can jointly perturb heterogeneous inputs, rendering unimodal certificates insufficient. We… ▽ More Randomized smoothing provides strong, model-agnostic robustness certificates, but existing guarantees are limited to single modalities, treating continuous and discrete inputs in isolation. This limitation becomes critical in multimodal models, where decisions depend on cross-modal semantics and adversaries can jointly perturb heterogeneous inputs, rendering unimodal certificates insufficient. We introduce a unified randomized smoothing framework for mixed discrete--continuous inputs based on an analytically tractable Neyman--Pearson formulation of the joint worst-case problem. By analyzing the joint likelihood ordering induced by factorized discrete and continuous noise, our approach yields a closed-form, one-dimensional certificate that strictly generalizes both Gaussian (image-only) and discrete (text-only) randomized smoothing. We validate the framework on multimodal safety filtering, providing, to our knowledge, the first model-agnostic Neyman--Pearson certificate for joint discrete-token and continuous-image perturbations in interaction-dependent text--image safety filtering. △ Less

Submitted 12 May, 2026; originally announced May 2026.

Comments: ICML 2026. Code: https://github.com/tdsai-lab/hybrid-randomized-smoothing

FraudBench: A Multimodal Benchmark for Detecting AI-Generated Fraudulent Refund Evidence

Authors: Xinyu Yan, Boyang Chen, Jiaming Zhang, Tiantong Wu, Hong Xi Tae, Yichen He, Tiantong Wang, Yachun Mi, Yurong Hao, Yilei Zhao, Lei Xiao, Longtao Huang, Pengjun Xie, Wei Liu, Wei Yang Bryan Lim

Abstract: Artificial Intelligence (AI)-generated images have become increasingly realistic and readily adaptable to concrete real-world claims, creating new challenges for verifying visual evidence. A concrete emerging risk is AI-generated refund fraud, in which manipulated or synthetic images are used to support claims about damaged products, poor delivery conditions, or service-related defects. Existing A… ▽ More Artificial Intelligence (AI)-generated images have become increasingly realistic and readily adaptable to concrete real-world claims, creating new challenges for verifying visual evidence. A concrete emerging risk is AI-generated refund fraud, in which manipulated or synthetic images are used to support claims about damaged products, poor delivery conditions, or service-related defects. Existing AI-generated image detection benchmarks mainly evaluate standalone authenticity classification, cross-generator transfer, or forensic localization, leaving claim-conditioned fraudulent evidence detection underexplored. To bridge this gap, we introduce FraudBench, a multimodal benchmark for detecting AI-generated fraudulent refund evidence. FraudBench is constructed from real-world user-review evidence across e-commerce, food delivery, and travel-service scenarios. We curate real evidence images together with their associated review and product metadata, identify genuine damaged and undamaged evidence through MLLM-assisted filtering and human annotation, and synthesize fake-damaged evidence from genuine undamaged reference images using six state-of-the-art image editing and generation models. Using FraudBench, we evaluate MLLMs, specialized AI-generated image detectors, and human participants under the same settings. Experiments show that current MLLMs often recognize real-damaged evidence but fail on many fake-damaged subsets, with fake-damage detection rates (TPR) far below the 50% baseline on most generator subsets. Specialized detectors generally perform better but remain inconsistent across generators and can produce false positives on real-damaged samples, revealing a clear gap between generic AI image detection and reliable claim-conditioned refund-evidence verification. △ Less

Submitted 9 May, 2026; originally announced May 2026.

Before Forgetting, Learn to Remember: Revisiting Foundational Learning Failures in LVLM Unlearning Benchmarks

Authors: JuneHyoung Kwon, MiHyeon Kim, Eunju Lee, JungMin Yun, Byeonggeuk Lim, YoungBin Kim

Abstract: While Large Vision-Language Models (LVLMs) offer powerful capabilities, they pose privacy risks by unintentionally memorizing sensitive personal information. Current unlearning benchmarks attempt to mitigate this using fictitious identities but overlook a critical stage 1 failure: models fail to effectively memorize target information initially, rendering subsequent unlearning evaluations unreliab… ▽ More While Large Vision-Language Models (LVLMs) offer powerful capabilities, they pose privacy risks by unintentionally memorizing sensitive personal information. Current unlearning benchmarks attempt to mitigate this using fictitious identities but overlook a critical stage 1 failure: models fail to effectively memorize target information initially, rendering subsequent unlearning evaluations unreliable. Diagnosing under-memorization and the multi-hop curse as root causes, we introduce ReMem, a Reliable Multi-hop and Multi-image Memorization Benchmark. ReMem ensures robust foundational learning through principled data scaling, reasoning-aware QA pairs, and diverse visual contexts. Additionally, we propose a novel Exposure metric to quantify the depth of information erasure from the model's internal probability distribution. Extensive experiments demonstrate that ReMem provides a rigorous and trustworthy framework for diagnosing both learning and unlearning behaviors in LVLMs. △ Less

Submitted 5 May, 2026; originally announced May 2026.

Comments: Accepted to Findings of ACL 2026

Tracing the Dynamics of Refusal: Exploiting Latent Refusal Trajectories for Robust Jailbreak Detection

Authors: Xulin Hu, Che Wang, Wei Yang Bryan Lim, Jianbo Gao, Zhong Chen

Abstract: Representation Engineering analyses often characterize refusal using static directions extracted from terminal or pooled representations. We ask whether this view misses how refusal is constructed across layer-token positions. Using causal tracing, we identify a \textit{Refusal Trajectory}: a sparse upstream activation pattern that often persists even when attacks such as GCG suppress terminal ref… ▽ More Representation Engineering analyses often characterize refusal using static directions extracted from terminal or pooled representations. We ask whether this view misses how refusal is constructed across layer-token positions. Using causal tracing, we identify a \textit{Refusal Trajectory}: a sparse upstream activation pattern that often persists even when attacks such as GCG suppress terminal refusal signals. Based on this observation, we propose SALO (Sparse Activation Localization Operator), a lightweight white-box detector that operates on raw hidden-state volumes from a selected layer window. Across Qwen, Llama, and Mistral models, SALO improves jailbreak detection on several attack families under a fixed XSTest-calibrated operating point. We further analyze static RepE-style baselines, ROI sensitivity, adaptive GCG attacks, and encoded-input boundary cases, clarifying both the promise and limitations of refusal-trajectory monitoring. △ Less

Submitted 26 May, 2026; v1 submitted 2 May, 2026; originally announced May 2026.

Comments: Accepted to the 43rd International Conference on Machine Learning (ICML 2026). Camera-ready version

CoAX: Cognitive-Oriented Attribution eXplanation User Model of Human Understanding of AI Explanations

Authors: Louth Bin Rawshan, Zhuoyu Wang, Brian Y. Lim

Abstract: Explainable AI (XAI) aims to improve user understanding and decisions when using AI models. However, despite innovations in XAI, recent user evaluations reveal that this goal remains elusive. Understanding human cognition can help explain why users struggle to effectively use AI explanations. Focusing on reasoning on structured (tabular) data, we examined various reasoning strategies for different… ▽ More Explainable AI (XAI) aims to improve user understanding and decisions when using AI models. However, despite innovations in XAI, recent user evaluations reveal that this goal remains elusive. Understanding human cognition can help explain why users struggle to effectively use AI explanations. Focusing on reasoning on structured (tabular) data, we examined various reasoning strategies for different XAI methods (none, feature importance, feature attribution) in the decision task of anticipating AI decisions (i.e., forward simulation). We i) elicited reasoning strategies from a formative user study, and ii) collected decisions from a summative user study. Using cognitive modeling, we implemented the processes underlying each reasoning strategy and evaluated their alignment with human decision-making. We found that our models better fit human decisions than baseline machine learning proxies, providing insights into which reasoning strategies are (in)effective. We then demonstrate how the fitted model can be used to form hypotheses and investigate research questions that are costly to study with real human participants. This work contributes to debugging human understanding of XAI, informing the future development of more usable and interpretable AI explanations. △ Less

Submitted 29 April, 2026; originally announced April 2026.

Aligning with Your Own Voice: Self-Corrected Preference Learning for Hallucination Mitigation in LVLMs

Authors: Byeonggeuk Lim, JungMin Yun, Junehyoung Kwon, Kyeonghyun Kim, YoungBin Kim

Abstract: Large Vision-Language Models (LVLMs) frequently suffer from hallucinations. Existing preference learning-based approaches largely rely on proprietary models to construct preference datasets. We identify that this reliance introduces a distributional mismatch between the proprietary and target models that hinders efficient alignment. To address this, we propose Alignment via VErified Self-correctio… ▽ More Large Vision-Language Models (LVLMs) frequently suffer from hallucinations. Existing preference learning-based approaches largely rely on proprietary models to construct preference datasets. We identify that this reliance introduces a distributional mismatch between the proprietary and target models that hinders efficient alignment. To address this, we propose Alignment via VErified Self-correction DPO (AVES-DPO), a framework that aligns LVLMs using in-distribution data derived from the model's intrinsic knowledge. Our approach employs a consensus-based verification mechanism to diagnose diverse hallucinations and guides the model to self-correct, thereby generating preference pairs strictly compatible with its internal distribution. Extensive experiments demonstrate that AVES-DPO surpasses existing baselines in hallucination mitigation while requiring only 5.2k samples. △ Less

Submitted 27 April, 2026; originally announced April 2026.

Comments: Accepted to ACL 2026

VG-CoT: Towards Trustworthy Visual Reasoning via Grounded Chain-of-Thought

Authors: Byeonggeuk Lim, Kyeonghyun Kim, JungMin Yun, YoungBin Kim

Abstract: The advancement of Large Vision-Language Models (LVLMs) requires precise local region-based reasoning that faithfully grounds the model's logic in actual visual evidence. However, existing datasets face limitations in scalability due to extensive manual annotation and lack of explicit alignment between multi-step reasoning and corresponding image regions, which constrains the evaluation of model t… ▽ More The advancement of Large Vision-Language Models (LVLMs) requires precise local region-based reasoning that faithfully grounds the model's logic in actual visual evidence. However, existing datasets face limitations in scalability due to extensive manual annotation and lack of explicit alignment between multi-step reasoning and corresponding image regions, which constrains the evaluation of model trustworthiness. To address these challenges, we propose the Visual Grounding Chain-of-Thought (VG-CoT) dataset, which explicitly links each reasoning step to real visual evidence within the image through a fully automated three-stage pipeline. The pipeline first extracts object- and text-level visual evidence using state-of-the-art detection and OCR models, then generates step-by-step grounded reasoning with GPT-4o, and finally refines the grounding through a rationale-driven open-set detection process. In addition, we introduce a new benchmark that comprehensively evaluates LVLMs reasoning across three complementary dimensions: Rationale Quality, Answer Accuracy, and Reasoning-Answer Alignment. Experiments with representative LVLMs, including LLaVA-1.5 and Qwen2-VL, demonstrate consistent improvements on most evaluation metrics, confirming that VG-CoT effectively enhances trustworthy, evidence-based reasoning while maintaining scalable and cost-efficient dataset construction. The dataset and code will be released publicly upon acceptance to facilitate further research. △ Less

Submitted 23 April, 2026; originally announced April 2026.

Comments: Accepted to LREC 2026

Quantifying Trust: Financial Risk Management for Trustworthy AI Agents

Authors: Wenyue Hua, Tianyi Peng, Chi Wang, Jiaxin Pei, Ian Kaufman, Bryan Lim, Chandler Fang

Abstract: Prior work on trustworthy AI emphasizes model-internal properties such as bias mitigation, adversarial robustness, and interpretability. As AI systems evolve into autonomous agents deployed in open environments and increasingly connected to payments or assets, the operational meaning of trust shifts to end-to-end outcomes: whether an agent completes tasks, follows user intent, and avoids failures… ▽ More Prior work on trustworthy AI emphasizes model-internal properties such as bias mitigation, adversarial robustness, and interpretability. As AI systems evolve into autonomous agents deployed in open environments and increasingly connected to payments or assets, the operational meaning of trust shifts to end-to-end outcomes: whether an agent completes tasks, follows user intent, and avoids failures that cause material or psychological harm. These risks are fundamentally product-level and cannot be eliminated by technical safeguards alone because agent behavior is inherently stochastic. To address this gap between model-level reliability and user-facing assurance, we propose a complementary framework based on risk management. Drawing inspiration from financial underwriting, we introduce the \textbf{Agentic Risk Standard (ARS)}, a payment settlement standard for AI-mediated transactions. ARS integrates risk assessment, underwriting, and compensation into a single transaction framework that protects users when interacting with agents. Under ARS, users receive predefined and contractually enforceable compensation in cases of execution failure, misalignment, or unintended outcomes. This shifts trust from an implicit expectation about model behavior to an explicit, measurable, and enforceable product guarantee. We also present a simulation study analyzing the social benefits of applying ARS to agentic transactions. ARS's implementation can be found at https://github.com/t54-labs/AgenticRiskStandard. △ Less

Submitted 4 May, 2026; v1 submitted 5 April, 2026; originally announced April 2026.

Comments: 30 pages, 9 figures

SafeGuard ASF: SR Agentic Humanoid Robot System for Autonomous Industrial Safety

Authors: Thanh Nguyen Canh, Thang Tran Viet, Thanh Tuan Tran, Ben Wei Lim

Abstract: The rise of unmanned ``dark factories'' operating without human presence demands autonomous safety systems capable of detecting and responding to multiple hazard types. We present SafeGuard ASF (Agentic Security Fleet), a comprehensive framework deploying humanoid robots for autonomous hazard detection in industrial environments. Our system integrates multi-modal perception (RGB-D imaging), a ReAc… ▽ More The rise of unmanned ``dark factories'' operating without human presence demands autonomous safety systems capable of detecting and responding to multiple hazard types. We present SafeGuard ASF (Agentic Security Fleet), a comprehensive framework deploying humanoid robots for autonomous hazard detection in industrial environments. Our system integrates multi-modal perception (RGB-D imaging), a ReAct-based agentic reasoning framework, and learned locomotion policies on the Unitree G1 humanoid platform. We address three critical hazard scenarios: fire and smoke detection, abnormal temperature monitoring in pipelines, and intruder detection in restricted zones. Our perception pipeline achieves 94.2% mAP for fire or smoke detection with 127ms latency. We train multiple locomotion policies, including dance motion tracking and velocity control, using Unitree RL Lab with PPO, demonstrating stable convergence within 80,000 training iterations. We validate our system in both simulation and real-world environments, demonstrating autonomous patrol, human detection with visual perception, and obstacle avoidance capabilities. The proposed ToolOrchestra action framework enables structured decision-making through perception, reasoning, and actuation tools. △ Less

Submitted 26 March, 2026; originally announced March 2026.

From Control to Foresight: Simulation as a New Paradigm for Human-Agent Collaboration

Authors: Gaole He, Brian Y. Lim

Abstract: Large Language Models (LLMs) are increasingly used to power autonomous agents for complex, multi-step tasks. However, human-agent interaction remains pointwise and reactive: users approve or correct individual actions to mitigate immediate risks, without visibility into subsequent consequences. This forces users to mentally simulate long-term effects, a cognitively demanding and often inaccurate p… ▽ More Large Language Models (LLMs) are increasingly used to power autonomous agents for complex, multi-step tasks. However, human-agent interaction remains pointwise and reactive: users approve or correct individual actions to mitigate immediate risks, without visibility into subsequent consequences. This forces users to mentally simulate long-term effects, a cognitively demanding and often inaccurate process. Users have control over individual steps but lack the foresight to make informed decisions. We argue that effective collaboration requires foresight, not just control. We propose simulation-in-the-loop, an interaction paradigm that enables users and agents to explore simulated future trajectories before committing to decisions. Simulation transforms intervention from reactive guesswork into informed exploration, while helping users discover latent constraints and preferences along the way. This perspective paper characterizes the limitations of current paradigms, introduces a conceptual framework for simulation-based collaboration, and illustrates its potential through concrete human-agent collaboration scenarios. △ Less

Submitted 12 March, 2026; originally announced March 2026.

Comments: CHI 2026 Workshop on Human-Agent Collaboration

Beyond Scores: Explainable Intelligent Assessment Strengthens Pre-service Teachers' Assessment Literacy

Authors: Yuang Wei, Fei Wang, Yifan Zhang, Brian Y. Lim, Bo Jiang

Abstract: Assessment literacy (AL) is essential for personalized education, yet difficult to cultivate in pre-service teachers. Conventional teacher preparation programs focus on theoretical knowledge, while digital assessment tools commonly provide opaque scores or parameters. These limitations hinder reflection and transfer, leaving AL underdeveloped. We propose XIA, an eXplainable Intelligent Assessment… ▽ More Assessment literacy (AL) is essential for personalized education, yet difficult to cultivate in pre-service teachers. Conventional teacher preparation programs focus on theoretical knowledge, while digital assessment tools commonly provide opaque scores or parameters. These limitations hinder reflection and transfer, leaving AL underdeveloped. We propose XIA, an eXplainable Intelligent Assessment platform that extends statistics-informed support with visualized cognitive diagnostic reasoning, including contrastive and counterfactual explanations. In a pre-post controlled study with 21 pre-service teachers, we combined quantitative tasks and questionnaires with qualitative interviews. The findings offer preliminary evidence that XIA supported reflection, self-regulation, and assessment awareness, and helped reduce assessment errors. Interviews further showed a shift from score-based judgments toward evidence-based reasoning. This work contributes insights into the design of intelligent assessment tools, showing how explanatory scaffolding can bridge assessment theory and classroom practice and support the cultivation of AL in teacher education. △ Less

Submitted 6 March, 2026; originally announced March 2026.

Comments: 26 pages,8 figures. Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI' 26)

Beyond Explainable AI (XAI): An Overdue Paradigm Shift and Post-XAI Research Directions

Authors: Saleh Afroogh, Syed Ishtiaque Ahmed, Petra Ahrweiler, David Alvarez-Melis, Mansur Maturidi Arief, Emilia Barakova, Falco J. Bargagli-Stoffi, Erdem Biyik, Hanjie Chen, Xiang 'Anthony' Chen, Robert Alan Clements, Keeley Crockett, Amit Dhurandhar, Fethiye Irmak Dogan, Mollie Dollinger, Motahhare Eslami, Aldo A Faisal, Arya Farahi, Melanie F. Pradier, Saadia Gabriel, Diego Garcia-Olano, Marzyeh Ghassemi, Shaona Ghosh, Hatice Gunes, Ehsan Hajiramezanali , et al. (24 additional authors not shown)

Abstract: This study provides a cross-disciplinary examination of Explainable Artificial Intelligence (XAI) approaches-focusing on deep neural networks (DNNs) and large language models (LLMs)-and identifies empirical and conceptual limitations in current XAI. We discuss critical symptoms that stem from deeper root causes (i.e., two paradoxes, two conceptual confusions, and five false assumptions). These fun… ▽ More This study provides a cross-disciplinary examination of Explainable Artificial Intelligence (XAI) approaches-focusing on deep neural networks (DNNs) and large language models (LLMs)-and identifies empirical and conceptual limitations in current XAI. We discuss critical symptoms that stem from deeper root causes (i.e., two paradoxes, two conceptual confusions, and five false assumptions). These fundamental problems within the current XAI research field reveal three insights: experimentally, XAI exhibits significant flaws; conceptually, it is paradoxical; and pragmatically, further attempts to reform the paradoxical XAI might exacerbate its confusion-demanding fundamental shifts and new research directions. To move beyond XAI's limitations, we propose a four-pronged synthesized paradigm shift toward reliable and certified AI development. These four components include: verification-focused Interactive AI (IAI) to establish scientific community protocols for certifying AI system performance rather than attempting post-hoc explanations, AI Epistemology for rigorous scientific foundations, User-Sensible AI to create context-aware systems tailored to specific user communities, and Model-Centered Interpretability for faithful technical analysis-together offering comprehensive post-XAI research directions. △ Less

Submitted 25 May, 2026; v1 submitted 27 February, 2026; originally announced February 2026.

MPU: Towards Secure and Privacy-Preserving Knowledge Unlearning for Large Language Models

Authors: Tiantong Wang, Xinyu Yan, Tiantong Wu, Yurong Hao, Pengjun Xie, Wei Yang Bryan Lim

Abstract: Machine unlearning for large language models often faces a privacy dilemma in which strict constraints prohibit sharing either the server's parameters or the client's forget set. To address this dual non-disclosure constraint, we propose MPU, an algorithm-agnostic privacy-preserving Multiple Perturbed Copies Unlearning framework that primarily introduces two server-side modules: Pre-Process for ra… ▽ More Machine unlearning for large language models often faces a privacy dilemma in which strict constraints prohibit sharing either the server's parameters or the client's forget set. To address this dual non-disclosure constraint, we propose MPU, an algorithm-agnostic privacy-preserving Multiple Perturbed Copies Unlearning framework that primarily introduces two server-side modules: Pre-Process for randomized copy generation and Post-Process for update aggregation. In Pre-Process, the server distributes multiple perturbed and reparameterized model instances, allowing the client to execute unlearning locally on its private forget set without accessing the server's exact original parameters. After local unlearning, the server performs Post-Process by inverting the reparameterization and aggregating updates with a harmonic denoising procedure to alleviate the impact of perturbation. Experiments with seven unlearning algorithms show that MPU achieves comparable unlearning performance to noise-free baselines, with most algorithms' average degradation well below 1% up to 10% noise, and can even outperform the noise-free baseline for some algorithms under 1% noise. Code is available at https://github.com/Tristan0318/MPU. △ Less

Submitted 14 May, 2026; v1 submitted 27 February, 2026; originally announced February 2026.

AdapTools: Adaptive Tool-based Indirect Prompt Injection Attacks on Agentic LLMs

Authors: Che Wang, Jiaming Zhang, Ziqi Zhang, Zijie Wang, Yinghui Wang, Jianbo Gao, Tao Wei, Zhong Chen, Wei Yang Bryan Lim

Abstract: The integration of external data services (e.g., Model Context Protocol, MCP) has made large language model-based agents increasingly powerful for complex task execution. However, this advancement introduces critical security vulnerabilities, particularly indirect prompt injection (IPI) attacks. Existing attack methods are limited by their reliance on static patterns and evaluation on simple langu… ▽ More The integration of external data services (e.g., Model Context Protocol, MCP) has made large language model-based agents increasingly powerful for complex task execution. However, this advancement introduces critical security vulnerabilities, particularly indirect prompt injection (IPI) attacks. Existing attack methods are limited by their reliance on static patterns and evaluation on simple language models, failing to address the fast-evolving nature of modern AI agents. We introduce AdapTools, a novel adaptive IPI attack framework that selects stealthier attack tools and generates adaptive attack prompts to create a rigorous security evaluation environment. Our approach comprises two key components: (1) Adaptive Attack Strategy Construction, which develops transferable adversarial strategies for prompt optimization, and (2) Attack Enhancement, which identifies stealthy tools capable of circumventing task-relevance defenses. Comprehensive experimental evaluation shows that AdapTools achieves a 2.13 times improvement in attack success rate while degrading system utility by a factor of 1.78. Notably, the framework maintains its effectiveness even against state-of-the-art defense mechanisms. Our method advances the understanding of IPI attacks and provides a useful reference for future research. △ Less

Submitted 24 February, 2026; originally announced February 2026.

Comments: 11 pages

ICON: Indirect Prompt Injection Defense for Agents based on Inference-Time Correction

Authors: Che Wang, Fuyao Zhang, Jiaming Zhang, Ziqi Zhang, Yinghui Wang, Longtao Huang, Jianbo Gao, Zhong Chen, Wei Yang Bryan Lim

Abstract: Large Language Model (LLM) agents are susceptible to Indirect Prompt Injection (IPI) attacks, where malicious instructions in retrieved content hijack the agent's execution. Existing defenses typically rely on strict filtering or refusal mechanisms, which suffer from a critical limitation: over-refusal, prematurely terminating valid agentic workflows. We propose ICON, a probing-to-mitigation frame… ▽ More Large Language Model (LLM) agents are susceptible to Indirect Prompt Injection (IPI) attacks, where malicious instructions in retrieved content hijack the agent's execution. Existing defenses typically rely on strict filtering or refusal mechanisms, which suffer from a critical limitation: over-refusal, prematurely terminating valid agentic workflows. We propose ICON, a probing-to-mitigation framework that neutralizes attacks while preserving task continuity. Our key insight is that IPI attacks leave distinct over-focusing signatures in the latent space. We introduce a Latent Space Trace Prober to detect attacks based on high intensity scores. Subsequently, a Mitigating Rectifier performs surgical attention steering that selectively manipulate adversarial query key dependencies while amplifying task relevant elements to restore the LLM's functional trajectory. Extensive evaluations on multiple backbones show that ICON achieves a competitive 0.4% ASR, matching commercial grade detectors, while yielding a over 50% task utility gain. Furthermore, ICON demonstrates robust Out of Distribution(OOD) generalization and extends effectively to multi-modal agents, establishing a superior balance between security and efficiency. △ Less

Submitted 24 February, 2026; originally announced February 2026.

Comments: 11 pages,

Rules or Weights? Comparing User Understanding of Explainable AI Techniques with the Cognitive XAI-Adaptive Model

Authors: Louth Bin Rawshan, Zhuoyu Wang, Brian Y Lim

Abstract: Rules and Weights are popular XAI techniques for explaining AI decisions. Yet, it remains unclear how to choose between them, lacking a cognitive framework to compare their interpretability. In an elicitation user study on forward and counterfactual decision tasks, we identified 7 reasoning strategies of interpreting three XAI Schemas - weights, rules, and their hybrid. To analyze their capabiliti… ▽ More Rules and Weights are popular XAI techniques for explaining AI decisions. Yet, it remains unclear how to choose between them, lacking a cognitive framework to compare their interpretability. In an elicitation user study on forward and counterfactual decision tasks, we identified 7 reasoning strategies of interpreting three XAI Schemas - weights, rules, and their hybrid. To analyze their capabilities, we propose CoXAM, a Cognitive XAI-Adaptive Model with shared memory representation to encode instance attributes, linear weights, and decision rules. CoXAM employs computational rationality to choose among reasoning processes based on the trade-off in utility and reasoning time, separately for forward or counterfactual decision tasks. In a validation study, CoXAM demonstrated a stronger alignment with human decision-making compared to baseline machine learning proxy models. The model successfully replicated and explained several key empirical findings, including that counterfactual tasks are inherently harder than forward tasks, decision tree rules are harder to recall and apply than linear weights, and the helpfulness of XAI depends on the application data context, alongside identifying which underlying reasoning strategies were most effective. With CoXAM, we contribute a cognitive basis to accelerate debugging and benchmarking disparate XAI techniques. △ Less

Submitted 23 February, 2026; originally announced February 2026.

Comparables XAI: Faithful Example-based AI Explanations with Counterfactual Trace Adjustments

Authors: Yifan Zhang, Tianle Ren, Fei Wang, Brian Y Lim

Abstract: Explaining with examples is an intuitive way to justify AI decisions. However, it is challenging to understand how a decision value should change relative to the examples with many features differing by large amounts. We draw from real estate valuation that uses Comparables-examples with known values for comparison. Estimates are made more accurate by hypothetically adjusting the attributes of eac… ▽ More Explaining with examples is an intuitive way to justify AI decisions. However, it is challenging to understand how a decision value should change relative to the examples with many features differing by large amounts. We draw from real estate valuation that uses Comparables-examples with known values for comparison. Estimates are made more accurate by hypothetically adjusting the attributes of each Comparable and correspondingly changing the value based on factors. We propose Comparables XAI for relatable example-based explanations of AI with Trace adjustments that trace counterfactual changes from each Comparable to the Subject, one attribute at a time, monotonically along the AI feature space. In modelling and user studies, Trace-adjusted Comparables achieved the highest XAI faithfulness and precision, user accuracy, and narrowest uncertainty bounds compared to linear regression, linearly adjusted Comparables, or unadjusted Comparables. This work contributes a new analytical basis for using example-based explanations to improve user understanding of AI decisions. △ Less

Submitted 14 February, 2026; originally announced February 2026.

Comments: Accepted by CHI 2026

Transferable XAI: Relating Understanding Across Domains with Explanation Transfer

Authors: Fei Wang, Yifan Zhang, Brian Y. Lim

Abstract: Current Explainable AI (XAI) focuses on explaining a single application, but when encountering related applications, users may rely on their prior understanding from previous explanations. This leads to either overgeneralization and AI overreliance, or burdensome independent memorization. Indeed, related decision tasks can share explanatory factors, but with some notable differences; e.g., body ma… ▽ More Current Explainable AI (XAI) focuses on explaining a single application, but when encountering related applications, users may rely on their prior understanding from previous explanations. This leads to either overgeneralization and AI overreliance, or burdensome independent memorization. Indeed, related decision tasks can share explanatory factors, but with some notable differences; e.g., body mass index (BMI) affects the risks for heart disease and diabetes at the same rate, but chest pain is more indicative of heart disease. Similarly, models using different attributes for the same task still share signals; e.g., temperature and pressure affect air pollution but in opposite directions due to the ideal gas law. Leveraging transfer of learning, we propose Transferable XAI to enable users to transfer understanding across related domains by explaining the relationship between domain explanations using a general affine transformation framework applied to linear factor explanations. The framework supports explanation transfer across various domain types: translation for data subspace (subsuming prior work on Incremental XAI), scaling for decision task, and mapping for attributes. Focusing on task and attributes domain types, in formative and summative user studies, we investigated how well participants could understand AI decisions from one domain to another. Compared to single-domain and domain-independent explanations, Transferable XAI was the most helpful for understanding the second domain, leading to the best decision faithfulness, factor recall, and ability to relate explanations between domains. This framework contributes to improving the reusability of explanations across related AI applications by explaining factor relationships between subspaces, tasks, and attributes. △ Less

Submitted 14 February, 2026; originally announced February 2026.

Comments: 40 pages, accepted by IUI2026

iRULER: Intelligible Rubric-Based User-Defined LLM Evaluation for Revision

Authors: Jingwen Bai, Wei Soon Cheong, Philippe Muller, Brian Y Lim

Abstract: Large Language Models (LLMs) have become indispensable for evaluating writing. However, text feedback they provide is often unintelligible, generic, and not specific to user criteria. Inspired by structured rubrics in education and intelligible AI explanations, we propose iRULER following identified design guidelines to \textit{scaffold} the review process by \textit{specific} criteria, providing… ▽ More Large Language Models (LLMs) have become indispensable for evaluating writing. However, text feedback they provide is often unintelligible, generic, and not specific to user criteria. Inspired by structured rubrics in education and intelligible AI explanations, we propose iRULER following identified design guidelines to \textit{scaffold} the review process by \textit{specific} criteria, providing \textit{justification} for score selection, and offering \textit{actionable} revisions to target different quality levels. To \textit{qualify} user-defined criteria, we recursively used iRULER with a rubric-of-rubrics to iteratively \textit{refine} rubrics. In controlled experiments on writing revision and rubric creation, iRULER most improved validated LLM-judged review scores and was perceived as most helpful and aligned compared to read-only rubric and text-based LLM feedback. Qualitative findings further support how iRULER satisfies the design guidelines for user-defined feedback. This work contributes interactive rubric tools for intelligible LLM-based review and revision of writing, and user-defined rubric creation. △ Less

Submitted 13 February, 2026; originally announced February 2026.

Comments: To Appear at CHI 2026

Editable XAI: Toward Bidirectional Human-AI Alignment with Co-Editable Explanations of Interpretable Attributes

Authors: Haoyang Chen, Jingwen Bai, Fang Tian, Brian Y Lim

Abstract: While Explainable AI (XAI) helps users understand AI decisions, misalignment in domain knowledge can lead to disagreement. This inconsistency hinders understanding, and because explanations are often read-only, users lack the control to improve alignment. We propose making XAI editable, allowing users to write rules to improve control and gain deeper understanding through the generation effect of… ▽ More While Explainable AI (XAI) helps users understand AI decisions, misalignment in domain knowledge can lead to disagreement. This inconsistency hinders understanding, and because explanations are often read-only, users lack the control to improve alignment. We propose making XAI editable, allowing users to write rules to improve control and gain deeper understanding through the generation effect of active learning. We developed CoExplain, leveraging a neural network for universal representation and symbolic rules for intuitive reasoning on interpretable attributes. CoExplain explains the neural network with a faithful proxy decision tree, parses user-written rules as an equivalent neural network graph, and collaboratively optimizes the decision tree. In a user study (N=43), CoExplain and manually editable XAI improved user understanding and model alignment compared to read-only XAI. CoExplain was easier to use with fewer edits and less time. This work contributes Editable XAI for bidirectional AI alignment, improving understanding and control. △ Less

Submitted 12 February, 2026; originally announced February 2026.

M-Loss: Quantifying Model Merging Compatibility with Limited Unlabeled Data

Authors: Tiantong Wang, Yiyang Duan, Haoyu Chen, Tiantong Wu, Wei Yang Bryan Lim

Abstract: Training of large-scale models is both computationally intensive and often constrained by the availability of labeled data. Model merging offers a compelling alternative by directly integrating the weights of multiple source models without requiring additional data or extensive training. However, conventional model merging techniques, such as parameter averaging, often suffer from the unintended c… ▽ More Training of large-scale models is both computationally intensive and often constrained by the availability of labeled data. Model merging offers a compelling alternative by directly integrating the weights of multiple source models without requiring additional data or extensive training. However, conventional model merging techniques, such as parameter averaging, often suffer from the unintended combination of non-generalizable features, especially when source models exhibit significant weight disparities. Comparatively, model ensembling generally provides more stable and superior performance that aggregates multiple models by averaging outputs. However, it incurs higher inference costs and increased storage requirements. While previous studies experimentally showed the similarities between model merging and ensembling, theoretical evidence and evaluation metrics remain lacking. To address this gap, we introduce Merging-ensembling loss (M-Loss), a novel evaluation metric that quantifies the compatibility of merging source models using very limited unlabeled data. By measuring the discrepancy between parameter averaging and model ensembling at layer and node levels, M-Loss facilitates more effective merging strategies. Specifically, M-Loss serves both as a quantitative criterion of the theoretical feasibility of model merging, and a guide for parameter significance in model pruning. Our theoretical analysis and empirical evaluations demonstrate that incorporating M-Loss into the merging process significantly improves the alignment between merged models and model ensembling, providing a scalable and efficient framework for accurate model consolidation. △ Less

Submitted 9 February, 2026; originally announced February 2026.

Comments: Code available at https://github.com/languangduan/mLoss

VENOMREC: Cross-Modal Interactive Poisoning for Targeted Promotion in Multimodal LLM Recommender Systems

Authors: Guowei Guan, Yurong Hao, Jiaming Zhang, Tiantong Wu, Fuyao Zhang, Tianxiang Chen, Longtao Huang, Cyril Leung, Wei Yang Bryan Lim

Abstract: Multimodal large language models (MLLMs) are pushing recommender systems (RecSys) toward content-grounded retrieval and ranking via cross-modal fusion. We find that while cross-modal consensus often mitigates conventional poisoning that manipulates interaction logs or perturbs a single modality, it also introduces a new attack surface where synchronised multimodal poisoning can reliably steer fuse… ▽ More Multimodal large language models (MLLMs) are pushing recommender systems (RecSys) toward content-grounded retrieval and ranking via cross-modal fusion. We find that while cross-modal consensus often mitigates conventional poisoning that manipulates interaction logs or perturbs a single modality, it also introduces a new attack surface where synchronised multimodal poisoning can reliably steer fused representations along stable semantic directions during fine-tuning. To characterise this threat, we formalise cross-modal interactive poisoning and propose VENOMREC, which performs Exposure Alignment to identify high-exposure regions in the joint embedding space and Cross-modal Interactive Perturbation to craft attention-guided coupled token--patch edits. Experiments on four real-world multimodal datasets demonstrate that VENOMREC consistently outperforms strong baselines, achieving 0.73 mean ER@20 and improving over the strongest baseline by +0.52 absolute ER points on average, while maintaining comparable recommendation utility. Code is available at https://github.com/GuoweiGuan666/VenomRec. △ Less

Submitted 21 July, 2026; v1 submitted 6 February, 2026; originally announced February 2026.

MarketGANs: Multivariate financial time-series data augmentation using generative adversarial networks

Authors: Jeonggyu Huh, Seungwon Jeong, Hyun-Gyoon Kim, Hyeng Keun Koo, Byung Hwa Lim

Abstract: This paper introduces MarketGAN, a factor-based generative framework for high-dimensional asset return generation under severe data scarcity. We embed an explicit asset-pricing factor structure as an economic inductive bias and generate returns as a single joint vector, thereby preserving cross-sectional dependence and tail co-movement alongside inter-temporal dynamics. MarketGAN employs generativ… ▽ More This paper introduces MarketGAN, a factor-based generative framework for high-dimensional asset return generation under severe data scarcity. We embed an explicit asset-pricing factor structure as an economic inductive bias and generate returns as a single joint vector, thereby preserving cross-sectional dependence and tail co-movement alongside inter-temporal dynamics. MarketGAN employs generative adversarial learning with a temporal convolutional network (TCN) backbone, which models stochastic, time-varying factor loadings and volatilities and captures long-range temporal dependence. Using daily returns of large U.S. equities, we find that MarketGAN more closely matches empirical stylized facts of asset returns, including heavy-tailed marginal distributions, volatility clustering, leverage effects, and, most notably, high-dimensional cross-sectional correlation structures and tail co-movement across assets, than conventional factor-model-based bootstrap approaches. In portfolio applications, covariance estimates derived from MarketGAN-generated samples outperform those derived from other methods when factor information is at least weakly informative, demonstrating tangible economic value. △ Less

Submitted 25 January, 2026; originally announced January 2026.

Advancing ESG Intelligence: An Expert-level Agent and Comprehensive Benchmark for Sustainable Finance

Authors: Yilei Zhao, Wentao Zhang, Lei Xiao, Yandan Zheng, Mengpu Liu, Wei Yang Bryan Lim

Abstract: Environmental, social, and governance (ESG) criteria are essential for evaluating corporate sustainability and ethical performance. However, professional ESG analysis is hindered by data fragmentation across unstructured sources, and existing large language models (LLMs) often struggle with the complex, multi-step workflows required for rigorous auditing. To address these limitations, we introduce… ▽ More Environmental, social, and governance (ESG) criteria are essential for evaluating corporate sustainability and ethical performance. However, professional ESG analysis is hindered by data fragmentation across unstructured sources, and existing large language models (LLMs) often struggle with the complex, multi-step workflows required for rigorous auditing. To address these limitations, we introduce ESGAgent, a hierarchical multi-agent system empowered by a specialized toolset, including retrieval augmentation, web search and domain-specific functions, to generate in-depth ESG analysis. Complementing this agentic system, we present a comprehensive three-level benchmark derived from 310 corporate sustainability reports, designed to evaluate capabilities ranging from atomic common-sense questions to the generation of integrated, in-depth analysis. Empirical evaluations demonstrate that ESGAgent outperforms state-of-the-art closed-source LLMs with an average accuracy of 84.15% on atomic question-answering tasks, and excels in professional report generation by integrating rich charts and verifiable references. These findings confirm the diagnostic value of our benchmark, establishing it as a vital testbed for assessing general and advanced agentic capabilities in high-stakes vertical domains. △ Less

Submitted 14 January, 2026; v1 submitted 13 January, 2026; originally announced January 2026.

Towards Unsupervised Causal Representation Learning via Latent Additive Noise Model Causal Autoencoders

Authors: Hans Jarett J. Ong, Brian Godwin S. Lim, Dominic Dayta, Renzo Roel P. Tan, Kazushi Ikeda

Abstract: Unsupervised representation learning seeks to recover latent generative factors, yet standard methods relying on statistical independence often fail to capture causal dependencies. A central challenge is identifiability: as established in disentangled representation learning and nonlinear ICA literature, disentangling causal variables from observational data is impossible without supervision, auxi… ▽ More Unsupervised representation learning seeks to recover latent generative factors, yet standard methods relying on statistical independence often fail to capture causal dependencies. A central challenge is identifiability: as established in disentangled representation learning and nonlinear ICA literature, disentangling causal variables from observational data is impossible without supervision, auxiliary signals, or strong inductive biases. In this work, we propose the Latent Additive Noise Model Causal Autoencoder (LANCA) to operationalize the Additive Noise Model (ANM) as a strong inductive bias for unsupervised discovery. Theoretically, we prove that while the ANM constraint does not guarantee unique identifiability in the general mixing case, it resolves component-wise indeterminacy by restricting the admissible transformations from arbitrary diffeomorphisms to the affine class. Methodologically, arguing that the stochastic encoding inherent to VAEs obscures the structural residuals required for latent causal discovery, LANCA employs a deterministic Wasserstein Auto-Encoder (WAE) coupled with a differentiable ANM Layer. This architecture transforms residual independence from a passive assumption into an explicit optimization objective. Empirically, LANCA outperforms state-of-the-art baselines on synthetic physics benchmarks (Pendulum, Flow), and on photorealistic environments (CANDLE), where it demonstrates superior robustness to spurious correlations arising from complex background scenes. △ Less

Submitted 15 December, 2025; originally announced December 2025.

Disrupting Hierarchical Reasoning: Adversarial Protection for Geographic Privacy in Multimodal Reasoning Models

Authors: Jiaming Zhang, Che Wang, Yang Cao, Longtao Huang, Wei Yang Bryan Lim

Abstract: Multi-modal large reasoning models (MLRMs) pose significant privacy risks by inferring precise geographic locations from personal images through hierarchical chain-of-thought reasoning. Existing privacy protection techniques, primarily designed for perception-based models, prove ineffective against MLRMs' sophisticated multi-step reasoning processes that analyze environmental cues. We introduce \t… ▽ More Multi-modal large reasoning models (MLRMs) pose significant privacy risks by inferring precise geographic locations from personal images through hierarchical chain-of-thought reasoning. Existing privacy protection techniques, primarily designed for perception-based models, prove ineffective against MLRMs' sophisticated multi-step reasoning processes that analyze environmental cues. We introduce \textbf{ReasonBreak}, a novel adversarial framework specifically designed to disrupt hierarchical reasoning in MLRMs through concept-aware perturbations. Our approach is founded on the key insight that effective disruption of geographic reasoning requires perturbations aligned with conceptual hierarchies rather than uniform noise. ReasonBreak strategically targets critical conceptual dependencies within reasoning chains, generating perturbations that invalidate specific inference steps and cascade through subsequent reasoning stages. To facilitate this approach, we contribute \textbf{GeoPrivacy-6K}, a comprehensive dataset comprising 6,341 ultra-high-resolution images ($\geq$2K) with hierarchical concept annotations. Extensive evaluation across seven state-of-the-art MLRMs (including GPT-o3, GPT-5, Gemini 2.5 Pro) demonstrates ReasonBreak's superior effectiveness, achieving a 14.4\% improvement in tract-level protection (33.8\% vs 19.4\%) and nearly doubling block-level protection (33.5\% vs 16.8\%). This work establishes a new paradigm for privacy protection against reasoning-based threats. △ Less

Submitted 29 March, 2026; v1 submitted 9 December, 2025; originally announced December 2025.

Comments: ICLR 2026

MultiPriv: Benchmarking Individual-Level Privacy Reasoning in Vision-Language Models

Authors: Xiongtao Sun, Hui Li, Jiaming Zhang, Yujie Yang, Kaili Liu, Ruxin Feng, Wen Jun Tan, Wei Yang Bryan Lim

Abstract: Modern Vision-Language Models (VLMs) pose significant individual-level privacy risks by linking fragmented multimodal data to identifiable individuals through hierarchical chain-of-thought reasoning. However, existing privacy benchmarks remain structurally insufficient for this threat, as they primarily evaluate privacy perception while failing to address the more critical risk of privacy reasonin… ▽ More Modern Vision-Language Models (VLMs) pose significant individual-level privacy risks by linking fragmented multimodal data to identifiable individuals through hierarchical chain-of-thought reasoning. However, existing privacy benchmarks remain structurally insufficient for this threat, as they primarily evaluate privacy perception while failing to address the more critical risk of privacy reasoning: a VLM's ability to infer and link distributed information to construct individual profiles. To address this gap, we propose MultiPriv, the first benchmark designed to systematically evaluate individual-level privacy reasoning in VLMs. We introduce the Privacy Perception and Reasoning (PPR) framework and construct a bilingual multimodal dataset with synthetic individual profiles, where identifiers, such as faces and names, are linked to sensitive attributes. This design enables nine challenging tasks spanning attribute detection, cross-image re-identification, and chained inference. We conduct a large-scale evaluation of over 50 open-source and commercial VLMs. In our controlled benchmark, 60% of widely used VLMs can perform individual-level privacy reasoning with up to 80% accuracy, suggesting a significant potential threat to personal privacy. The benchmark is available at https://github.com/CyberChangAn/MultiPriv-PII. △ Less

Submitted 29 May, 2026; v1 submitted 20 November, 2025; originally announced November 2025.

Communication-Pipelined Split Federated Learning for Foundation Model Fine-Tuning in UAV Networks

Authors: Zizhen Zhou, Ying-Chang Liang, Yanyu Cheng, Wei Yang Bryan Lim

Abstract: Deploying foundation models (FMs) on uncrewed aerial vehicles (UAVs) promises broad ``low-altitude economy'' applications. Split federated learning (SFL)-based fine-tuning leverages distributed data while keeping raw data local and reduces client-side burden by partitioning the model between client and server. However, the per-round training latency is dominated by stragglers. Training paradigms f… ▽ More Deploying foundation models (FMs) on uncrewed aerial vehicles (UAVs) promises broad ``low-altitude economy'' applications. Split federated learning (SFL)-based fine-tuning leverages distributed data while keeping raw data local and reduces client-side burden by partitioning the model between client and server. However, the per-round training latency is dominated by stragglers. Training paradigms featuring parallel gradient transmission (GT) allocate dedicated portions of downlink communication resources to each client. They may leave resources idle and suffer from prolonged GT latency, especially in UAV networks, where the communication latency typically far exceeds the computation latency. To address this, we propose a sequential GT paradigm, where the server dedicates all downlink resources for the current GT. We further propose communication-pipelined SFL (CPSFL), characterized by downlink GT priority scheduling and intra-round asynchronous training. We investigate CPSFL-based LoRA fine-tuning of FMs in UAV networks and formulate an optimization problem to minimize a weighted sum of per-round training latency and worst-case client energy consumption by optimizing the split point selection (SPS) and the computing and communication resource allocation (CCRA) (the uplink bandwidth allocation and the server computing frequency allocation). To solve this problem, we develop an attention-based deep reinforcement learning (DRL) framework, where the base station agent decides the split point and the CCRA in each round by leveraging previous round information, including UAV trajectories. Simulation results show that the proposed DRL-based CPSFL scheme outperforms the parallel GT benchmarks, the ablation variants, the fixed CCRA scheme, while approaching the best fixed-SPS scheme. △ Less

Submitted 19 November, 2025; originally announced November 2025.

Comments: This work has been submitted to the IEEE for possible publication

DualTAP: A Dual-Task Adversarial Protector for Mobile MLLM Agents

Authors: Fuyao Zhang, Jiaming Zhang, Che Wang, Xiongtao Sun, Yurong Hao, Guowei Guan, Wenjie Li, Longtao Huang, Wei Yang Bryan Lim

Abstract: The reliance of mobile GUI agents on Multimodal Large Language Models (MLLMs) introduces a severe privacy vulnerability: screenshots containing Personally Identifiable Information (PII) are often sent to untrusted, third-party routers. These routers can exploit their own MLLMs to mine this data, violating user privacy. Existing privacy perturbations fail the critical dual challenge of this scenari… ▽ More The reliance of mobile GUI agents on Multimodal Large Language Models (MLLMs) introduces a severe privacy vulnerability: screenshots containing Personally Identifiable Information (PII) are often sent to untrusted, third-party routers. These routers can exploit their own MLLMs to mine this data, violating user privacy. Existing privacy perturbations fail the critical dual challenge of this scenario: protecting PII from the router's MLLM while simultaneously preserving task utility for the agent's MLLM. To address this gap, we propose the Dual-Task Adversarial Protector (DualTAP), a novel framework that, for the first time, explicitly decouples these conflicting objectives. DualTAP trains a lightweight generator using two key innovations: (i) a contrastive attention module that precisely identifies and targets only the PII-sensitive regions, and (ii) a dual-task adversarial objective that simultaneously minimizes a task-preservation loss (to maintain agent utility) and a privacy-interference loss (to suppress PII leakage). To facilitate this study, we introduce PrivScreen, a new dataset of annotated mobile screenshots designed specifically for this dual-task evaluation. Comprehensive experiments on six diverse MLLMs (e.g., GPT-5) demonstrate DualTAP's state-of-the-art protection. It reduces the average privacy leakage rate by 31.6 percentage points (a 3.0x relative improvement) while, critically, maintaining an 80.8% task success rate - a negligible drop from the 83.6% unprotected baseline. DualTAP presents the first viable solution to the privacy-utility trade-off in mobile MLLM agents. △ Less

Submitted 17 November, 2025; originally announced November 2025.

Enhancing Graph Representations with Neighborhood-Contextualized Message-Passing

Authors: Brian Godwin Lim, Galvin Brice Lim, Renzo Roel Tan, Irwin King, Kazushi Ikeda

Abstract: Graph neural networks (GNNs) have become an indispensable tool for analyzing relational data. Classical GNNs are broadly classified into three variants: convolutional, attentional, and message-passing. While the standard message-passing variant is expressive, its typical pair-wise messages only consider the features of the center node and each neighboring node individually. This design fails to in… ▽ More Graph neural networks (GNNs) have become an indispensable tool for analyzing relational data. Classical GNNs are broadly classified into three variants: convolutional, attentional, and message-passing. While the standard message-passing variant is expressive, its typical pair-wise messages only consider the features of the center node and each neighboring node individually. This design fails to incorporate contextual information contained within the broader local neighborhood, potentially hindering its ability to learn meaningful relationships within the entire set of neighboring nodes. To address this, the paper first refines the concept of neighborhood-contextualization within GNNs, leveraging ideas from set-based aggregation methods and a key property of the attentional variant. This then serves as the basis for generalizing the message-passing variant to the proposed neighborhood-contextualized message-passing (NCMP) framework. To demonstrate its utility, a simple, mathematically grounded method to parametrize and operationalize NCMP is presented, leading to the development of the proposed Soft-Isomorphic Neighborhood-Contextualized Graph Convolution Network (SINC-GCN). Across a diverse set of synthetic and benchmark datasets, SINC-GCN strikes a highly favorable balance between expressivity and efficiency. Notably, while more complex models incur significant computational overhead, SINC-GCN delivers substantial performance gains with considerable effect sizes over baseline GNN models while maintaining a highly efficient asymptotic runtime complexity, further underscoring the distinctive utility of neighborhood-contextualization. Overall, by integrating multiset neighborhood context, the proposed NCMP framework serves as a practical and scalable path toward enhancing the graph representational power of classical GNNs. △ Less

Submitted 30 June, 2026; v1 submitted 14 November, 2025; originally announced November 2025.

Comments: Published in Transactions on Machine Learning Research

Journal ref: Transactions on Machine Learning Research. (2026)

A Shared-Autonomy Construction Robotic System for Overhead Works

Authors: David Minkwan Kim, K. M. Brian Lee, Yong Hyeok Seo, Nikola Raicevic, Runfa Blark Li, Kehan Long, Chan Seon Yoon, Dong Min Kang, Byeong Jo Lim, Young Pyoung Kim, Nikolay Atanasov, Truong Nguyen, Se Woong Jun, Young Wook Kim

Abstract: We present the ongoing development of a robotic system for overhead work such as ceiling drilling. The hardware platform comprises a mobile base with a two-stage lift, on which a bimanual torso is mounted with a custom-designed drilling end effector and RGB-D cameras. To support teleoperation in dynamic environments with limited visibility, we use Gaussian splatting for online 3D reconstruction an… ▽ More We present the ongoing development of a robotic system for overhead work such as ceiling drilling. The hardware platform comprises a mobile base with a two-stage lift, on which a bimanual torso is mounted with a custom-designed drilling end effector and RGB-D cameras. To support teleoperation in dynamic environments with limited visibility, we use Gaussian splatting for online 3D reconstruction and introduce motion parameters to model moving objects. For safe operation around dynamic obstacles, we developed a neural configuration-space barrier approach for planning and control. Initial feasibility studies demonstrate the capability of the hardware in drilling, bolting, and anchoring, and the software in safe teleoperation in a dynamic environment. △ Less

Submitted 12 November, 2025; originally announced November 2025.

Comments: 4pages, 8 figures, ICRA construction workshop

Using Language Models as Closed-Loop High-Level Planners for Robotics Applications: A Brief Overview and Benchmarks

Authors: Hao Wang, Sathwik Karnik, Bea Lim, Somil Bansal

Abstract: Large Language Models (LLMs) and Vision Language Models (VLMs) have become popular tools for embodied high-level planning. However, their deployment in black-box settings often leads to unpredictable or costly errors. To harness their capabilities more reliably in robotic systems, we empirically investigate practical strategies for integrating language models as closed-loop planners. Concretely, w… ▽ More Large Language Models (LLMs) and Vision Language Models (VLMs) have become popular tools for embodied high-level planning. However, their deployment in black-box settings often leads to unpredictable or costly errors. To harness their capabilities more reliably in robotic systems, we empirically investigate practical strategies for integrating language models as closed-loop planners. Concretely, we study how the control horizon and warm-starting impact the performance of language model-based planners. We design and conduct controlled experiments to extract actionable insights, providing recommendations that can help improve the performance and robustness of language model-based embodied planning. The full implementation and experiments are available on the project website △ Less

Submitted 26 April, 2026; v1 submitted 10 November, 2025; originally announced November 2025.

Dynamic Factor Analysis of Price Movements in the Philippine Stock Exchange

Authors: Brian Godwin Lim, Dominic Dayta, Benedict Ryan Tiu, Renzo Roel Tan, Len Patrick Dominic Garces, Kazushi Ikeda

Abstract: The intricate dynamics of stock markets have led to extensive research on models that are able to effectively explain their inherent complexities. This study leverages the econometrics literature to explore the dynamic factor model as an interpretable model with sufficient predictive capabilities for capturing essential market phenomena. Although the model has been extensively applied for predicti… ▽ More The intricate dynamics of stock markets have led to extensive research on models that are able to effectively explain their inherent complexities. This study leverages the econometrics literature to explore the dynamic factor model as an interpretable model with sufficient predictive capabilities for capturing essential market phenomena. Although the model has been extensively applied for predictive purposes, this study focuses on analyzing the extracted loadings and common factors as an alternative framework for understanding stock price dynamics. The results reveal novel insights into traditional market theories when applied to the Philippine Stock Exchange using the Kalman method and maximum likelihood estimation, with subsequent validation against the capital asset pricing model. Notably, a one-factor model extracts a common factor representing systematic or market dynamics similar to the composite index, whereas a two-factor model extracts common factors representing market trends and volatility. Furthermore, an application of the model for nowcasting the growth rates of the Philippine gross domestic product highlights the potential of the extracted common factors as viable real-time market indicators, yielding over a 34% decrease in the out-of-sample prediction error. Overall, the results underscore the value of dynamic factor analysis in gaining a deeper understanding of market price movement dynamics. △ Less

Submitted 8 October, 2025; originally announced October 2025.

Journal ref: Financial Innovation. 12(2026)

Graph of Agents: Principled Long Context Modeling by Emergent Multi-Agent Collaboration

Authors: Taejong Joo, Shu Ishida, Ivan Sosnovik, Bryan Lim, Sahand Rezaei-Shoshtari, Adam Gaier, Robert Giaquinto

Abstract: As a model-agnostic approach to long context modeling, multi-agent systems can process inputs longer than a large language model's context window without retraining or architectural modifications. However, their performance often heavily relies on hand-crafted multi-agent collaboration strategies and prompt engineering, which limit generalizability. In this work, we introduce a principled framewor… ▽ More As a model-agnostic approach to long context modeling, multi-agent systems can process inputs longer than a large language model's context window without retraining or architectural modifications. However, their performance often heavily relies on hand-crafted multi-agent collaboration strategies and prompt engineering, which limit generalizability. In this work, we introduce a principled framework that formalizes the model-agnostic long context modeling problem as a compression problem, yielding an information-theoretic compression objective. Building on this framework, we propose Graph of Agents (GoA), which dynamically constructs an input-dependent collaboration structure that maximizes this objective. For Llama 3.1 8B and Qwen3 8B across six document question answering benchmarks, GoA improves the average $F_1$ score of retrieval-augmented generation by 5.7\% and a strong multi-agent baseline using a fixed collaboration structure by 16.35\%, respectively. Even with only a 2K context window, GoA surpasses the 128K context window Llama 3.1 8B on LongBench, showing a dramatic increase in effective context length. Our source code is available at https://github.com/tjoo512/graph-of-agents. △ Less

Submitted 26 September, 2025; originally announced September 2025.

Comments: Preprint

Enhancing XAI Interpretation through a Reverse Mapping from Insights to Visualizations

Authors: Aniket Nuthalapati, Nicholas Hinds, Brian Y. Lim, Qianwen Wang

Abstract: As AI systems become increasingly integrated into high-stakes domains, enabling users to accurately interpret model behavior is critical. While AI explanations can be provided, users often struggle to reason effectively with these explanations, limiting their ability to validate or learn from AI decisions. To address this gap, we introduce Reverse Mapping, a novel approach that enhances visual exp… ▽ More As AI systems become increasingly integrated into high-stakes domains, enabling users to accurately interpret model behavior is critical. While AI explanations can be provided, users often struggle to reason effectively with these explanations, limiting their ability to validate or learn from AI decisions. To address this gap, we introduce Reverse Mapping, a novel approach that enhances visual explanations by incorporating user-derived insights back into the explanation workflow. Our system extracts structured insights from free-form user interpretations using a large language model and maps them back onto visual explanations through interactive annotations and coordinated multi-view visualizations. Inspired by the verification loop in the visualization knowledge generation model, this design aims to foster more deliberate, reflective interaction with AI explanations. We demonstrate our approach in a prototype system with two use cases and qualitative user feedback. △ Less

Submitted 25 August, 2025; originally announced August 2025.

Comments: 5 pages, 5 figures, accepted by IEEE VIS 2025

Oblivionis: A Lightweight Learning and Unlearning Framework for Federated Large Language Models

Authors: Fuyao Zhang, Xinyu Yan, Tiantong Wu, Wenjie Li, Tianxiang Chen, Yang Cao, Ran Yan, Longtao Huang, Wei Yang Bryan Lim, Qiang Yang

Abstract: Large Language Models (LLMs) increasingly leverage Federated Learning (FL) to utilize private, task-specific datasets for fine-tuning while preserving data privacy. However, while federated LLM frameworks effectively enable collaborative training without raw data sharing, they critically lack built-in mechanisms for regulatory compliance like GDPR's right to be forgotten. Integrating private data… ▽ More Large Language Models (LLMs) increasingly leverage Federated Learning (FL) to utilize private, task-specific datasets for fine-tuning while preserving data privacy. However, while federated LLM frameworks effectively enable collaborative training without raw data sharing, they critically lack built-in mechanisms for regulatory compliance like GDPR's right to be forgotten. Integrating private data heightens concerns over data quality and long-term governance, yet existing distributed training frameworks offer no principled way to selectively remove specific client contributions post-training. Due to distributed data silos, stringent privacy constraints, and the intricacies of interdependent model aggregation, federated LLM unlearning is significantly more complex than centralized LLM unlearning. To address this gap, we introduce Oblivionis, a lightweight learning and unlearning framework that enables clients to selectively remove specific private data during federated LLM training, enhancing trustworthiness and regulatory compliance. By unifying FL and unlearning as a dual optimization objective, we incorporate 6 FL and 5 unlearning algorithms for comprehensive evaluation and comparative analysis, establishing a robust pipeline for federated LLM unlearning. Extensive experiments demonstrate that Oblivionis outperforms local training, achieving a robust balance between forgetting efficacy and model utility, with cross-algorithm comparisons providing clear directions for future LLM development. △ Less

Submitted 8 November, 2025; v1 submitted 12 August, 2025; originally announced August 2025.

Q-CLIP: Unleashing the Power of Vision-Language Models for Video Quality Assessment through Unified Cross-Modal Adaptation

Authors: Yachun Mi, Yu Li, Yanting Li, Chen Hui, Tong Zhang, Zhixuan Li, Chenyue Song, Wei Yang Bryan Lim, Shaohui Liu

Abstract: Accurate and efficient Video Quality Assessment (VQA) has long been a key research challenge. Current mainstream VQA methods typically improve performance by pretraining on large-scale classification datasets (e.g., ImageNet, Kinetics-400), followed by fine-tuning on VQA datasets. However, this strategy presents two significant challenges: (1) merely transferring semantic knowledge learned from pr… ▽ More Accurate and efficient Video Quality Assessment (VQA) has long been a key research challenge. Current mainstream VQA methods typically improve performance by pretraining on large-scale classification datasets (e.g., ImageNet, Kinetics-400), followed by fine-tuning on VQA datasets. However, this strategy presents two significant challenges: (1) merely transferring semantic knowledge learned from pretraining is insufficient for VQA, as video quality depends on multiple factors (e.g., semantics, distortion, motion, aesthetics); (2) pretraining on large-scale datasets demands enormous computational resources, often dozens or even hundreds of times greater than training directly on VQA datasets. Recently, Vision-Language Models (VLMs) have shown remarkable generalization capabilities across a wide range of visual tasks, and have begun to demonstrate promising potential in quality assessment. In this work, we propose Q-CLIP, the first fully VLMs-based framework for VQA. Q-CLIP enhances both visual and textual representations through a Shared Cross-Modal Adapter (SCMA), which contains only a minimal number of trainable parameters and is the only component that requires training. This design significantly reduces computational cost. In addition, we introduce a set of five learnable quality-level prompts to guide the VLMs in perceiving subtle quality variations, thereby further enhancing the model's sensitivity to video quality. Furthermore, we investigate the impact of different frame sampling strategies on VQA performance, and find that frame-difference-based sampling leads to better generalization performance across datasets. Extensive experiments demonstrate that Q-CLIP exhibits excellent performance on several VQA datasets. △ Less

Submitted 9 October, 2025; v1 submitted 8 August, 2025; originally announced August 2025.

Spattack: Subgroup Poisoning Attacks on Federated Recommender Systems

Authors: Bo Yan, Yurong Hao, Dingqi Liu, Huabin Sun, Pengpeng Qiao, Wei Yang Bryan Lim, Yang Cao, Chuan Shi

Abstract: Federated recommender systems (FedRec) have emerged as a promising approach to provide personalized recommendations while protecting user privacy. However, recent studies have shown their vulnerability to poisoning attacks, where malicious clients inject crafted gradients to promote target items to benign users. Existing attacks typically target the full user group, which compromises stealth and i… ▽ More Federated recommender systems (FedRec) have emerged as a promising approach to provide personalized recommendations while protecting user privacy. However, recent studies have shown their vulnerability to poisoning attacks, where malicious clients inject crafted gradients to promote target items to benign users. Existing attacks typically target the full user group, which compromises stealth and increases detection risk. In contrast, real-world adversaries may prefer to target specific user subgroups, such as promoting health supplements to older individuals, to maximize effectiveness while preserving stealth. Motivated by this gap, we introduce Spattack, the first poisoning attack designed to manipulate recommendations for specific user subgroups in federated settings. Spattack adopts an approximate-and-promote paradigm, which approximates user embeddings of target and non-target subgroups and then promotes target items to the target subgroup. We further reveal a trade-off between strong attack performance on the target subgroup and limited impact on the non-target subgroup. To achieve a better trade-off, we propose enhanced approximation and promotion strategies. For approximation, we push embeddings of different subgroups apart via contrastive learning and augment the target subgroup's relevant item set through clustering. For promotion, we align embeddings of target items and relevant items to strengthen their semantic connections, together with an adaptive weighting strategy to balance effects across subgroups. Experiments on three real-world datasets demonstrate that Spattack achieves strong attack performance on the target subgroup with minimal impact on non-target users, even when only 0.1% of users are malicious. Moreover, Spattack maintains competitive recommendation performance and shows strong resilience against mainstream defenses. △ Less

Submitted 30 January, 2026; v1 submitted 7 July, 2025; originally announced July 2025.

Comments: Accepted by WWW 2026

CAVALRY-V: A Large-Scale Generator Framework for Adversarial Attacks on Video MLLMs

Authors: Jiaming Zhang, Rui Hu, Qing Guo, Wei Yang Bryan Lim

Abstract: Video Multimodal Large Language Models (V-MLLMs) have shown impressive capabilities in temporal reasoning and cross-modal understanding, yet their vulnerability to adversarial attacks remains underexplored due to unique challenges: complex cross-modal reasoning mechanisms, temporal dependencies, and computational constraints. We present CAVALRY-V (Cross-modal Language-Vision Adversarial Yielding f… ▽ More Video Multimodal Large Language Models (V-MLLMs) have shown impressive capabilities in temporal reasoning and cross-modal understanding, yet their vulnerability to adversarial attacks remains underexplored due to unique challenges: complex cross-modal reasoning mechanisms, temporal dependencies, and computational constraints. We present CAVALRY-V (Cross-modal Language-Vision Adversarial Yielding for Videos), a novel framework that directly targets the critical interface between visual perception and language generation in V-MLLMs. Our approach introduces two key innovations: (1) a dual-objective semantic-visual loss function that simultaneously disrupts the model's text generation logits and visual representations to undermine cross-modal integration, and (2) a computationally efficient two-stage generator framework that combines large-scale pre-training for cross-model transferability with specialized fine-tuning for spatiotemporal coherence. Empirical evaluation on comprehensive video understanding benchmarks demonstrates that CAVALRY-V significantly outperforms existing attack methods, achieving 22.8% average improvement over the best baseline attacks on both commercial systems (GPT-4.1, Gemini 2.0) and open-source models (QwenVL-2.5, InternVL-2.5, Llava-Video, Aria, MiniCPM-o-2.6). Our framework achieves flexibility through implicit temporal coherence modeling rather than explicit regularization, enabling significant performance improvements even on image understanding (34.4% average gain). This capability demonstrates CAVALRY-V's potential as a foundational approach for adversarial research across multimodal systems. △ Less

Submitted 1 July, 2025; originally announced July 2025.