In the age of agentic AI, the growing deployment of multi-modal models (MMs) has introduced new attack vectors that can leak sensitive training data in MMs, causing privacy leakage. This paper investigates a black-box privacy attack, i.e., membership inference attack (MIA) on multi-modal vision-language models (VLMs). State-of-the-art research analyzes privacy attacks primarily to unimodal AI-ML systems, while recent studies indicate MMs can also be vulnerable to privacy attacks. While researchers have demonstrated that biologically inspired neural network representations can improve unimodal model resilience against adversarial attacks, it remains unexplored whether neuro-inspired MMs are resilient against privacy attacks. In this work, we introduce a systematic neuroscience-inspired topological regularization (tau) framework to analyze MM VLMs resilience against image-text-based inference privacy attacks. We examine this phenomenon using three VLMs: BLIP, PaliGemma 2, and ViT-GPT2, across three benchmark datasets: COCO, CC3M, and NoCaps. Our experiments compare the resilience of baseline and neuro VLMs (with topological regularization), where the tau > 0 configuration defines the NEURO variant of VLM. Our results on the BLIP model using the COCO dataset illustrate that MIA attack success in NEURO VLMs drops by 24% mean ROC-AUC, while achieving similar model utility (similarities between generated and reference captions) in terms of MPNet and ROUGE-2 metrics. This shows neuro VLMs are comparatively more resilient against privacy attacks, while not significantly compromising model utility. Our extensive evaluation with PaliGemma 2 and ViT-GPT2 models, on two additional datasets: CC3M and NoCaps, further validates the consistency of the findings. This work contributes to the growing understanding of privacy risks in MMs and provides evidence on neuro VLMs privacy threat resilience.

Large language models (LLMs) and autonomous coding agents are increasingly used to generate software across a wide range of domains. Yet a core requirement remains unmet: ensuring that generated code is secure without compromising its functional correctness. Existing benchmarks and evaluations for secure code generation fall short-many measure only vulnerability reduction, disregard correctness preservation, or evaluate security and functionality on separate datasets, violating the fundamental need for simultaneous joint evaluation. We present DUALGAUGE, the first fully automated benchmarking framework designed to rigorously evaluate the security and correctness of LLM-generated code in unison. Given the lack of datasets enabling joint evaluation of secure code generation, we also present DUALGAUGE-BENCH, a curated benchmark suite of diverse coding tasks, each paired with manually validated test suites for both security and functionality, designed for full coverage of specification requirements. At the core of DUALGAUGE is an agentic program executor, which runs a program against given tests in sandboxed environments, and an LLM-based evaluator, which assesses both correctness and vulnerability behavior against expected outcomes. We rigorously evaluated and ensured the quality of DUALGAUGE-BENCH and the accuracy of DUALGAUGE, and applied DUALGAUGE to benchmarking ten leading LLMs on DUALGAUGE-BENCH across thousands of test scenarios. Our results reveal critical gaps in correct and secure code generation by these LLMs, for which our open-source system and datasets help accelerate progress via reproducible, scalable, and rigorous evaluation.

While recent work has extended CoT to multimodal settings, achieving state-of-the-art results on science question answering benchmarks like ScienceQA, the generalizabil-ity of these approaches across diverse domains remains un-derexplored. This work presents a comprehensive analysis of Multimodal Chain-of-Thought (Multimodal-CoT) reasoning, evaluating its effectiveness on the A-OKVQA, OKVQA and ChartQA datasets, which requires broad commonsense and world knowledge beyond scientific reasoning. We implement the two-stage framework proposed by Zhang et al. [3], which separates rationale generation from answer inference and integrates vision features through a gated fusion mechanism with T5-based language models. Through systematic ablation studies, we analyze the contributions of vision features, rationale quality, and architectural choices. Our findings reveal that while vision integration significantly reduces hallucination in rationale generation, the effectiveness of CoT reasoning varies substantially across question types, with commonsense reasoning presenting particular challenges. This work provides practical insights for researchers implementing multimodal reasoning systems and identifies key areas for future improvement in cross-domain generalization.

Scientific reasoning through Large Language Models in heliophysics involves more than just recalling facts: it requires incorporating physical assumptions, maintaining consistent units, and providing clear scientific formats through coordinated approaches. To address these challenges, we present Reasoning With a Star, a newly contributed heliophysics dataset applicable to reasoning; we also provide an initial benchmarking approach. Our data are constructed from National Aeronautics and Space Administration & University Corporation for Atmospheric Research Living With a Star summer school problem sets and compiled into a readily consumable question-and-answer structure with question contexts, reasoning steps, expected answer type, ground-truth targets, format hints, and metadata. A programmatic grader checks the predictions using unit-aware numerical tolerance, symbolic equivalence, and schema validation. We benchmark a single-shot baseline and four multi-agent patterns, finding that decomposing workflows through systems engineering principles outperforms direct prompting on problems requiring deductive reasoning rather than pure inductive recall.

Large language models (LLMs) have shown strong potential in automating the design of agentic workflows. However, existing methods still rely heavily on manually predefined operators, limiting generalization and scalability. To address this issue, we propose $A^2Flow$, a fully automated framework for agentic workflow generation based on self-adaptive abstraction operators. $A^2Flow$ employs a three-stage operator extraction process: 1) Case-based Initial Operator Generation: leveraging expert demonstrations and LLM reasoning to generate case-specific operators; 2) Operator Clustering and Preliminary Abstraction: grouping similar operators across tasks to form preliminary abstractions; and 3) Deep Extraction for Abstract Execution Operators: applying long chain-of-thought prompting and multi-path reasoning to derive compact and generalizable execution operators. These operators serve as reusable building blocks for workflow construction without manual predefinition. Furthermore, we enhance node-level workflow search with an operator memory mechanism, which retains historical outputs to enrich context and improve decision-making. Experiments on general and embodied benchmarks show that $A^2Flow$ achieves a 2.4\% and 19.3\% average performance improvement and reduces resource usage by 37\% over state-of-the-art baselines. Homepage:https://github.com/pandawei-ele/A2FLOW

Large language models (LLMs) increasingly mediate human decision-making and behaviour. Ensuring LLM processing of moral meaning therefore has become a critical challenge. Current approaches rely predominantly on bottom-up methods such as fine-tuning and reinforcement learning from human feedback. We propose a fundamentally different approach: embedding moral meaning processing directly into the architectural mechanisms and frameworks of transformer-based models through top-down design principles. We first sketch a framework that conceptualizes attention as a dynamic interface mediating between structure and processing, contrasting with existing linear attention frameworks in psychology. We start from established biological-artificial attention analogies in neural architecture design to improve cognitive processing. We extend this analysis to moral processing, using Iris Murdoch's theory of loving attention (sustained, just observation that enables moral transformation by reseeing others with clarity and compassion) to philosophically discuss functional analogies between human and LLM moral processing. We formulate and evaluate potentially promising technical operationalizations to embed morality in LLM architectures and frameworks. We acknowledge the limitations of our exploration and give three key contributions. (1) We conceptualize attention as a dynamic system mechanism mediating between structure and processing. (2) Drawing on the Murdoch notion of loving attention, we outline technical pathways for embedding morality in LLMs, through modified training objectives, runtime weight adjustments, and architectural refinements to attention. (3) We argue that integrating morality into architectures and frameworks complements external, constraint-based methods. We conclude with a call for collaboration between transformer designers and philosophers engaged in AI ethics.

The rapid evolution of generative AI necessitates robust safety evaluations. However, current safety datasets are predominantly English-centric, failing to capture specific risks in non-English, socio-cultural contexts such as Korean, and are often limited to the text modality. To address this gap, we introduce AssurAI, a new quality-controlled Korean multimodal dataset for evaluating the safety of generative AI. First, we define a taxonomy of 35 distinct AI risk factors, adapted from established frameworks by a multidisciplinary expert group to cover both universal harms and relevance to the Korean socio-cultural context. Second, leveraging this taxonomy, we construct and release AssurAI, a large-scale Korean multimodal dataset comprising 11,480 instances across text, image, video, and audio. Third, we apply the rigorous quality control process used to ensure data integrity, featuring a two-phase construction (i.e., expert-led seeding and crowdsourced scaling), triple independent annotation, and an iterative expert red-teaming loop. Our pilot study validates AssurAI's effectiveness in assessing the safety of recent LLMs. We release AssurAI to the public to facilitate the development of safer and more reliable generative AI systems for the Korean community.

Contemporary large language model deployments typically employ uniform prompting strategies across diverse query types, applying verbose response patterns to both complex analytical tasks and straightforward factual questions. This one-size-fits-all methodology leads to substantial token inefficiency, a concern amplified by the significant cost differential between input and output tokens--the latter commanding 4-8x higher prices across major providers. We present Dynamic Template Selection (DTS), which adaptively matches response templates to query complexity, achieving significant cost reductions without compromising response quality. We compared two routing approaches: a simple MLP that uses pre-computed embeddings and a more complex fine-tuned RoBERTa transformer. Through comprehensive evaluation on 1,000 MMLU questions, we find that the MLP router achieves 90.5% routing accuracy on held-out test data, marginally exceeding RoBERTa's performance (89.5%) despite utilizing 125M fewer parameters. Notably, our empirical analysis reveals provider-agnostic behavior in template selection--routing decisions generalize effectively across 3 major LLM providers (OpenAI GPT-4, Google Gemini, and Anthropic Claude), as validated through 9,000 production API calls. While routing accuracy remains consistent at 90.5% across providers, observed token reductions vary from 32.6% to 33.9%, reflecting provider-specific generation characteristics. This work contributes several key elements: formal problem formulation with theoretical grounding in machine learning, four algorithms with corresponding complexity analyses, and extensive empirical validation across production systems.

Despite high performance on clinical benchmarks, large language models may reach correct conclusions through faulty reasoning, a failure mode with safety implications for oncology decision support that is not captured by accuracy-based evaluation. In this two-cohort retrospective study, we developed a hierarchical taxonomy of reasoning errors from GPT-4 chain-of-thought responses to real oncology notes and tested its clinical relevance. Using breast and pancreatic cancer notes from the CORAL dataset, we annotated 600 reasoning traces to define a three-tier taxonomy mapping computational failures to cognitive bias frameworks. We validated the taxonomy on 822 responses from prostate cancer consult notes spanning localized through metastatic disease, simulating extraction, analysis, and clinical recommendation tasks. Reasoning errors occurred in 23 percent of interpretations and dominated overall errors, with confirmation bias and anchoring bias most common. Reasoning failures were associated with guideline-discordant and potentially harmful recommendations, particularly in advanced disease management. Automated evaluators using state-of-the-art language models detected error presence but could not reliably classify subtypes. These findings show that large language models may provide fluent but clinically unsafe recommendations when reasoning is flawed. The taxonomy provides a generalizable framework for evaluating and improving reasoning fidelity before clinical deployment.

Hyperbolic geometry is an effective geometry for embedding hierarchical data structures. Hyperbolic learning has therefore become increasingly prominent in machine learning applications where data is hierarchically organized or governed by hierarchical semantics, ranging from recommendation systems to computer vision. The quality of hyperbolic embeddings is tightly coupled to the structure of the input hierarchy, which is often derived from knowledge graphs or ontologies. Recent work has uncovered that for an optimal hyperbolic embedding, a high branching factor and single inheritance are key, while embedding algorithms are robust to imbalance and hierarchy size. To assist knowledge engineers in reorganizing hierarchical knowledge, this paper investigates whether Large Language Models (LLMs) have the ability to automatically restructure hierarchies to meet these criteria. We propose a prompt-based approach to transform existing hierarchies using LLMs, guided by known desiderata for hyperbolic embeddings. Experiments on 16 diverse hierarchies show that LLM-restructured hierarchies consistently yield higher-quality hyperbolic embeddings across several standard embedding quality metrics. Moreover, we show how LLM-guided hierarchy restructuring enables explainable reorganizations, providing justifications to knowledge engineers.