Ensuring LLM alignment is critical to information security as AI models become increasingly widespread and integrated in society. Unfortunately, many defenses against adversarial attacks and jailbreaking on LLMs cannot adapt quickly to new attacks, degrade model responses to benign prompts, or introduce significant barriers to scalable implementation. To mitigate these challenges, we introduce a real-time, self-tuning (RTST) moderator framework to defend against adversarial attacks while maintaining a lightweight training footprint. We empirically evaluate its effectiveness using Google's Gemini models against modern, effective jailbreaks. Our results demonstrate the advantages of an adaptive, minimally intrusive framework for jailbreak defense over traditional fine-tuning or classifier models.

--Cardiovascular disease (CVD) prediction remains a tremendous challenge due to its multifactorial etiology and global burden of morbidity and mortality. Despite the growing availability of genomic and electrophysiological data, extracting biologically meaningful insights from such high-dimensional, noisy, and sparsely annotated datasets remains a non-trivial task. Recently, LLMs has been applied effectively to predict structural variations in biological sequences. In this work, we explore the potential of fine-tuned LLMs to predict cardiac diseases and SNPs potentially leading to CVD risk using genetic markers derived from high-throughput genomic profiling. We investigate the effect of genetic patterns associated with cardiac conditions and evaluate how LLMs can learn latent biological relationships from structured and semi-structured genomic data obtained by mapping genetic aspects that are inherited from the family tree. By framing the problem as a Chain of Thought (CoT) reasoning task, the models are prompted to generate disease labels and articulate informed clinical deductions across diverse patient profiles and phenotypes. The findings highlight the promise of LLMs in contributing to early detection, risk assessment, and ultimately, the advancement of personalized medicine in cardiac care. Cardiovascular disease (CVDs) remain the most common cause of mortality globally. In 2023, approximately 20.5 million people died from CVDs [1], accounting for about one-third of all global deaths. This marks a significant increase from the estimated 17.9 million CVD deaths in 2019.

Graph Neural Networks (GNNs) have emerged as powerful tools for learning over structured data, including text-attributed graphs, which are common in domains such as citation networks, social platforms, and knowledge graphs. GNNs are not inherently interpretable and thus, many explanation methods have been proposed. However, existing explanation methods often struggle to generate interpretable, fine-grained rationales, especially when node attributes include rich natural language. In this work, we introduce LOGIC, a lightweight, post-hoc framework that uses large language models (LLMs) to generate faithful and interpretable explanations for GNN predictions. LOGIC projects GNN node embeddings into the LLM embedding space and constructs hybrid prompts that interleave soft prompts with textual inputs from the graph structure. This enables the LLM to reason about GNN internal representations and produce natural language explanations along with concise explanation subgraphs. Our experiments across four real-world TAG datasets demonstrate that LOGIC achieves a favorable trade-off between fidelity and sparsity, while significantly improving human-centric metrics such as insightfulness. LOGIC sets a new direction for LLM-based explainability in graph learning by aligning GNN internals with human reasoning.

Large language models (LLMs) have achieved impressive performance, leading to their widespread adoption as decision-support tools in resource-constrained contexts like hiring and admissions. There is, however, scientific consensus that AI systems can reflect and exacerbate societal biases, raising concerns about identity-based harm when used in critical social contexts. Prior work has laid a solid foundation for assessing bias in LLMs by evaluating demographic disparities in different language reasoning tasks. In this work, we extend single-axis fairness evaluations to examine intersectional bias, recognizing that when multiple axes of discrimination intersect, they create distinct patterns of disadvantage. We create a new benchmark called WinoIdentity by augmenting the WinoBias dataset with 25 demographic markers across 10 attributes, including age, nationality, and race, intersected with binary gender, yielding 245,700 prompts to evaluate 50 distinct bias patterns. Focusing on harms of omission due to underrepresentation, we investigate bias through the lens of uncertainty and propose a group (un)fairness metric called Coreference Confidence Disparity which measures whether models are more or less confident for some intersectional identities than others. We evaluate five recently published LLMs and find confidence disparities as high as 40% along various demographic attributes including body type, sexual orientation and socio-economic status, with models being most uncertain about doubly-disadvantaged identities in anti-stereotypical settings. Surprisingly, coreference confidence decreases even for hegemonic or privileged markers, indicating that the recent impressive performance of LLMs is more likely due to memorization than logical reasoning. Notably, these are two independent failures in value alignment and validity that can compound to cause social harm.

Large reasoning models achieve strong performance through test-time scaling but incur substantial computational overhead, particularly from excessive token generation when processing short input prompts. While sparse attention mechanisms can reduce latency and memory usage, existing approaches suffer from significant accuracy degradation due to accumulated errors during long-generation reasoning. These methods generally require either high token retention rates or expensive retraining. We introduce LessIsMore, a training-free sparse attention mechanism for reasoning tasks, which leverages global attention patterns rather than relying on traditional head-specific local optimizations. LessIsMore aggregates token selections from local attention heads with recent contextual information, enabling unified cross-head token ranking for future decoding layers. This unified selection improves generalization and efficiency by avoiding the need to maintain separate token subsets per head. Evaluation across diverse reasoning tasks and benchmarks shows that LessIsMore preserves -- and in some cases improves -- accuracy while achieving a $1.1\times$ average decoding speed-up compared to full attention. Moreover, LessIsMore attends to $2\times$ fewer tokens without accuracy loss, achieving a $1.13\times$ end-to-end speed-up compared to existing sparse attention methods.

Large language models are known to produce outputs that are plausible but factually incorrect. To prevent people from making erroneous decisions by blindly trusting AI, researchers have explored various ways of communicating factuality estimates in AI-generated outputs to end-users. However, little is known about whether revealing content estimated to be factually incorrect influences users' trust when compared to hiding it altogether. We tested four different ways of disclosing an AI-generated output with factuality assessments: transparent (highlights less factual content), attention (highlights factual content), opaque (removes less factual content), ambiguity (makes less factual content vague), and compared them with a baseline response without factuality information. We conducted a human subjects research (N = 148) using the strategies in question-answering scenarios. We found that the opaque and ambiguity strategies led to higher trust while maintaining perceived answer quality, compared to the other strategies. We discuss the efficacy of hiding presumably less factual content to build end-user trust.

Evaluating social biases in language models (LMs) is crucial for ensuring fairness and minimizing the reinforcement of harmful stereotypes in AI systems. Existing benchmarks, such as the Bias Benchmark for Question Answering (BBQ), primarily focus on Western contexts, limiting their applicability to the Indian context. To address this gap, we introduce BharatBBQ, a culturally adapted benchmark designed to assess biases in Hindi, English, Marathi, Bengali, Tamil, Telugu, Odia, and Assamese. BharatBBQ covers 13 social categories, including 3 intersectional groups, reflecting prevalent biases in the Indian sociocultural landscape. Our dataset contains 49,108 examples in one language that are expanded using translation and verification to 392,864 examples in eight different languages. We evaluate five multilingual LM families across zero and few-shot settings, analyzing their bias and stereotypical bias scores. Our findings highlight persistent biases across languages and social categories and often amplified biases in Indian languages compared to English, demonstrating the necessity of linguistically and culturally grounded benchmarks for bias evaluation.

We introduce SQL-Exchange, a framework for mapping SQL queries across different database schemas by preserving the source query structure while adapting domain-specific elements to align with the target schema. We investigate the conditions under which such mappings are feasible and beneficial, and examine their impact on enhancing the in-context learning performance of text-to-SQL systems as a downstream task. Our comprehensive evaluation across multiple model families and benchmark datasets--assessing structural alignment with source queries, execution validity on target databases, and semantic correctness--demonstrates that SQL-Exchange is effective across a wide range of schemas and query types. Our results further show that using mapped queries as in-context examples consistently improves text-to-SQL performance over using queries from the source schema.

Large Language Models (LLMs) are increasingly applied to medical imaging tasks, including image interpretation and synthetic image generation. However, these models often produce hallucinations, which are confident but incorrect outputs that can mislead clinical decisions. This study examines hallucinations in two directions: image to text, where LLMs generate reports from X-ray, CT, or MRI scans, and text to image, where models create medical images from clinical prompts. We analyze errors such as factual inconsistencies and anatomical inaccuracies, evaluating outputs using expert informed criteria across imaging modalities. Our findings reveal common patterns of hallucination in both interpretive and generative tasks, with implications for clinical reliability. We also discuss factors contributing to these failures, including model architecture and training data. By systematically studying both image understanding and generation, this work provides insights into improving the safety and trustworthiness of LLM driven medical imaging systems.

Text-to-SQL bridges the gap between natural language and structured database language, thus allowing non-technical users to easily query databases. Traditional approaches model text-to-SQL as a direct translation task, where a given Natural Language Query (NLQ) is mapped to an SQL command. Recent advances in large language models (LLMs) have significantly improved translation accuracy, however, these methods all require that the target database is pre-specified. This becomes problematic in scenarios with multiple extensive databases, where identifying the correct database becomes a crucial yet overlooked step. In this paper, we propose a three-stage end-to-end text-to-SQL framework to identify the user's intended database before generating SQL queries. Our approach leverages LLMs and prompt engineering to extract implicit information from natural language queries (NLQs) in the form of a ruleset. We then train a large db\_id prediction model, which includes a RoBERTa-based finetuned encoder, to predict the correct Database identifier (db\_id) based on both the NLQ and the LLM-generated rules. Finally, we refine the generated SQL by using critic agents to correct errors. Experimental results demonstrate that our framework outperforms the current state-of-the-art models in both database intent prediction and SQL generation accuracy.