The paper introduces VisAnomReasoner, a parameter-efficient Vision-Language Model (VLM), trained on a new benchmark (VisAnomBench) to accurately and interpretably detect anomalies in time-series data.

The paper introduces FraudBench, a multimodal benchmark designed to detect AI-generated fraudulent refund evidence, finding that current AI models struggle significantly with claim-conditioned fake-da…

The paper proposes a novel Text-Guided Backdoor (TGB) attack that uses common words in text descriptions as stealthy triggers for multimodal models, enhancing practicality and controllability.

The paper proposes an end-to-end, deployable blueprint for an in-line machine-vision system that not only inspects carpet defects in real-time but also systematically collects and labels defect data t…

KidsNanny is a two-stage multimodal content moderation pipeline that achieves high accuracy and efficiency in detecting child safety threats, particularly excelling in text-embedded content.

The paper analyzes that while multimodal large language models (MLLMs) offer superior semantic understanding for image generation, this enhanced capability significantly increases safety risks, partic…

The paper introduces Auto-ART, a comprehensive open-source framework that provides structured meta-analysis and automated testing for adversarial robustness, revealing significant gaps in current ML s…

The paper introduces a robust, two-part framework (HyPE and HyPS) using hyperbolic geometry to efficiently detect and sanitize malicious prompts targeting Vision-Language Models (VLMs).

The paper introduces a dual-dimension evaluation for universal adversarial attacks on Vision-Language Models (VLMs), demonstrating that high reported attack success rates significantly overestimate th…

This study systematically evaluates Vision Mamba models for detecting AI-generated images, finding that while they show promise, their current strengths and limitations must be understood relative to…

The study demonstrates that robust, domain-invariant representations of synthetic deception can be rapidly entrenched in LLMs using modest fine-tuning, detectable by linear probes even in early layers…

The paper explains the 'table-chart gap' in scientific claim verification by showing that multimodal LLMs successfully encode information from charts but fail to route it to the final prediction layer…

The paper addresses 'Template Collapse' in 3D CT report generation—where models generate generic reports—by proposing CLarGen, a decoupled framework that significantly improves clinical accuracy and d…

The paper introduces Responsible Contrastive Soft Prompting (RCSP), a parameter-efficient method using soft prompts to improve LLM reliability by simultaneously suppressing hallucinations, encouraging…

This study demonstrates that Chain-of-Thought (CoT) monitoring is fundamentally fragile and unreliable for detecting misaligned behavior across typologically diverse languages, especially in low-resou…

The paper introduces OpenClawBench, a large-scale dataset and framework for measuring process-side anomalies in real-world agent execution trajectories, demonstrating that task success does not guaran…

The paper introduces ImageProtector, a user-side method that embeds an imperceptible perturbation into images to prevent Multi-modal Large Language Models (MLLMs) from analyzing and extracting sensiti…

This paper systematically diagnoses the failure modes of linear deception probes in LLMs, finding that while single-direction probes are insufficient, multi-dimensional probes can recover robust detec…

The paper introduces OmniMatBench, a comprehensive, human-calibrated multimodal reasoning benchmark covering 19 materials science subfields, revealing that current multimodal language models (MLLMs) h…