Clinical tasks such as diagnosis and treatment require strong decision-making abilities, highlighting the importance of rigorous evaluation benchmarks to assess the reliability of large language models (LLMs). In this work, we introduce a knowledge-guided data augmentation framework that enhances the difficulty of clinical multiple-choice question (MCQ) datasets by generating distractors (i.e., incorrect choices that are similar to the correct one and may confuse existing LLMs). Using our KG-based pipeline, the generated choices are both clinically plausible and deliberately misleading. Our approach involves multi-step, semantically informed walks on a medical knowledge graph to identify distractor paths-associations that are medically relevant but factually incorrect-which then guide the LLM in crafting more deceptive distractors. We apply the designed knowledge graph guided distractor generation (KGGDG) pipline, to six widely used medical QA benchmarks and show that it consistently reduces the accuracy of state-of-the-art LLMs. These findings establish KGGDG as a powerful tool for enabling more robust and diagnostic evaluations of medical LLMs.

Recent work has demonstrated the plausibility of frontier AI models scheming -- knowingly and covertly pursuing an objective misaligned with its developer's intentions. Such behavior could be very hard to detect, and if present in future advanced systems, could pose severe loss of control risk. It is therefore important for AI developers to rule out harm from scheming prior to model deployment. In this paper, we present a suite of scheming reasoning evaluations measuring two types of reasoning capabilities that we believe are prerequisites for successful scheming: First, we propose five evaluations of ability to reason about and circumvent oversight (stealth). Second, we present eleven evaluations for measuring a model's ability to instrumentally reason about itself, its environment and its deployment (situational awareness). We demonstrate how these evaluations can be used as part of a scheming inability safety case: a model that does not succeed on these evaluations is almost certainly incapable of causing severe harm via scheming in real deployment. We run our evaluations on current frontier models and find that none of them show concerning levels of either situational awareness or stealth.

Agentic search such as Deep Research systems-where agents autonomously browse the web, synthesize information, and return comprehensive citation-backed answers-represents a major shift in how users interact with web-scale information. While promising greater efficiency and cognitive offloading, the growing complexity and open-endedness of agentic search have outpaced existing evaluation benchmarks and methodologies, which largely assume short search horizons and static answers. In this paper, we introduce Mind2Web 2, a benchmark of 130 realistic, high-quality, and long-horizon tasks that require real-time web browsing and extensive information synthesis, constructed with over 1000 hours of human labor. To address the challenge of evaluating time-varying and complex answers, we propose a novel Agent-as-a-Judge framework. Our method constructs task-specific judge agents based on a tree-structured rubric design to automatically assess both answer correctness and source attribution. We conduct a comprehensive evaluation of ten frontier agentic search systems and human performance, along with a detailed error analysis to draw insights for future development. The best-performing system, OpenAI Deep Research, can already achieve 50-70% of human performance while spending half the time, highlighting its great potential. Altogether, Mind2Web 2 provides a rigorous foundation for developing and benchmarking the next generation of agentic search systems.

Large language models (LLMs) have democratized software development, reducing the expertise barrier for programming complex applications. This accessibility extends to malicious software development, raising significant security concerns. While LLM providers have implemented alignment mechanisms to prevent direct generation of overtly malicious code, these safeguards predominantly evaluate individual prompts in isolation, overlooking a critical vulnerability: malicious operations can be systematically decomposed into benign-appearing sub-tasks. In this paper, we introduce the Malware Generation Compiler (MGC), a novel framework that leverages this vulnerability through modular decomposition and alignment-evasive generation. MGC employs a specialized Malware Description Intermediate Representation (MDIR) to bridge high-level malicious intents and benign-appearing code snippets. Extensive evaluation demonstrates that our attack reliably generates functional malware across diverse task specifications and categories, outperforming jailbreaking methods by +365.79% and underground services by +78.07% in correctness on three benchmark datasets. Case studies further show that MGC can reproduce and even enhance 16 real-world malware samples. This work provides critical insights for security researchers by exposing the risks of compositional attacks against aligned AI systems. Demonstrations are available at https://sites.google.com/view/malware-generation-compiler.

Keyword decision in Sponsored Search Advertising is critical to the success of ad campaigns. While LLM-based methods offer automated keyword generation, they face three major limitations: reliance on large-scale query-keyword pair data, lack of online multi-objective performance monitoring and optimization, and weak quality control in keyword selection. These issues hinder the agentic use of LLMs in fully automating keyword decisions by monitoring and reasoning over key performance indicators such as impressions, clicks, conversions, and CTA effectiveness. To overcome these challenges, we propose OMS, a keyword generation framework that is On-the-fly (requires no training data, monitors online performance, and adapts accordingly), Multi-objective (employs agentic reasoning to optimize keywords based on multiple performance metrics), and Self-reflective (agentically evaluates keyword quality). Experiments on benchmarks and real-world ad campaigns show that OMS outperforms existing methods; ablation and human evaluations confirm the effectiveness of each component and the quality of generated keywords.

Background: The use of Electronic Health Records (EHRs) for epidemiological studies and artificial intelligence (AI) training is increasing rapidly. The reliability of the results depends on the accuracy and completeness of EHR data. However, EHR data often contain significant quality issues, including misrepresentations of subpopulations, biases, and systematic errors, as they are primarily collected for clinical and billing purposes. Existing quality assessment methods remain insufficient, lacking systematic procedures to assess data fitness for research. Methods: We present the Medical Data Pecking approach, which adapts unit testing and coverage concepts from software engineering to identify data quality concerns. We demonstrate our approach using the Medical Data Pecking Tool (MDPT), which consists of two main components: (1) an automated test generator that uses large language models and grounding techniques to create a test suite from data and study descriptions, and (2) a data testing framework that executes these tests, reporting potential errors and coverage. Results: We evaluated MDPT on three datasets: All of Us (AoU), MIMIC-III, and SyntheticMass, generating 55-73 tests per cohort across four conditions. These tests correctly identified 20-43 non-aligned or non-conforming data issues. We present a detailed analysis of the LLM-generated test suites in terms of reference grounding and value accuracy. Conclusion: Our approach incorporates external medical knowledge to enable context-sensitive data quality testing as part of the data analysis workflow to improve the validity of its outcomes. Our approach tackles these challenges from a quality assurance perspective, laying the foundation for further development such as additional data modalities and improved grounding methods.

Event stream based scene text recognition is a newly arising research topic in recent years which performs better than the widely used RGB cameras in extremely challenging scenarios, especially the low illumination, fast motion. Existing works either adopt end-to-end encoder-decoder framework or large language models for enhanced recognition, however, they are still limited by the challenges of insufficient interpretability and weak contextual logical reasoning. In this work, we propose a novel chain-of-thought reasoning based event stream scene text recognition framework, termed ESTR-CoT. Specifically, we first adopt the vision encoder EVA-CLIP (ViT-G/14) to transform the input event stream into tokens and utilize a Llama tokenizer to encode the given generation prompt. A Q-former is used to align the vision token to the pre-trained large language model Vicuna-7B and output both the answer and chain-of-thought (CoT) reasoning process simultaneously. Our framework can be optimized using supervised fine-tuning in an end-to-end manner. In addition, we also propose a large-scale CoT dataset to train our framework via a three stage processing (i.e., generation, polish, and expert verification). This dataset provides a solid data foundation for the development of subsequent reasoning-based large models. Extensive experiments on three event stream STR benchmark datasets (i.e., EventSTR, WordArt*, IC15*) fully validated the effectiveness and interpretability of our proposed framework. The source code and pre-trained models will be released on https://github.com/Event-AHU/ESTR-CoT.

--We introduce DeST A2.5-Audio, a general-purpose Large Audio Language Model (LALM) designed for robust auditory perception and instruction-following, without requiring task-specific audio instruction-tuning. Recent LALMs typically augment Large Language Models (LLMs) with auditory capabilities by training on large-scale, manually curated or LLM-synthesized audio-instruction datasets. However, these approaches have often suffered from the catastrophic forgetting of the LLM's original language abilities. T o address this, we revisit the data construction pipeline and propose DeST A, a self-generated cross-modal alignment strategy in which the backbone LLM generates its own training targets. This approach preserves the LLM's native language proficiency while establishing effective audio-text alignment, thereby enabling zero-shot generalization without task-specific tuning. Using DeST A, we construct DeST A-AQA5M, a large-scale, task-agnostic dataset containing 5 million training samples derived from 7,000 hours of audio spanning 50 diverse datasets, including speech, environmental sounds, and music. DeST A2.5-Audio achieves state-of-the-art or competitive performance across a wide range of audio-language benchmarks, including Dynamic-SUPERB, MMAU, SAKURA, Speech-IFEval, and V oiceBench. Comprehensive comparative studies demonstrate that our self-generated strategy outperforms widely adopted data construction and training strategies in both auditory perception and instruction-following capabilities. Our findings underscore the importance of carefully designed data construction in LALM development and offer practical insights for building robust, general-purpose LALMs. HE development of general-purpose artificial intelligence has become a central focus in contemporary AI research, driven by the remarkable performance of large language models (LLMs) across various natural language understanding and generation tasks [1]-[7]. Building on these advancements, a promising direction is to equip LLMs with multi-modal understanding capabilities, leading to the emergence of Large Audio Language Models (LALMs) [8]-[22] and Large Vision Language Models (L VLMs) [23]-[27]. This paper focuses on building a general-purpose LALM, illustrated in Figure 1. To develop a general-purpose LALM, two core capabilities are essential: auditory perception and instruction-following. Auditory perception refers to the comprehensive processing of auditory information, including speech, non-verbal cues, background sounds, and music.

The rise of Large Language Models (LLMs) has enabled the development of specialized AI agents with domain-specific reasoning and interaction capabilities, particularly in healthcare. While recent frameworks simulate medical decision-making, they largely focus on single-turn tasks where a doctor agent receives full case information upfront -- diverging from the real-world diagnostic process, which is inherently uncertain, interactive, and iterative. In this paper, we introduce MIMIC-Patient, a structured dataset built from the MIMIC-III electronic health records (EHRs), designed to support dynamic, patient-level simulations. Building on this, we propose DynamiCare, a novel dynamic multi-agent framework that models clinical diagnosis as a multi-round, interactive loop, where a team of specialist agents iteratively queries the patient system, integrates new information, and dynamically adapts its composition and strategy. We demonstrate the feasibility and effectiveness of DynamiCare through extensive experiments, establishing the first benchmark for dynamic clinical decision-making with LLM-powered agents.

As one of their many applications, large language models (LLMs) have recently shown promise in automating register transfer level (RTL) code generation. However, conventional LLM decoding strategies, originally designed for natural language, often fail to meet the structural and semantic demands of RTL, leading to hallucinated, repetitive, or invalid code outputs. In this paper, we first investigate the root causes of these decoding failures through an empirical analysis of token-level entropy during RTL generation. Our findings reveal that LLMs exhibit low confidence in regions of structural ambiguity or semantic complexity, showing that standard decoding strategies fail to differentiate between regions requiring determinism (syntax-critical regions) and those that benefit from creative exploratory variability (design-critical regions). Then, to overcome this, we introduce DecoRTL, a novel run-time decoding strategy, that is both syntax-aware and contrastive for RTL code generation. DecoRTL integrates two complementary components: (i) self-consistency sampling, which generates multiple candidates and re-ranks them based on token-level agreement to promote correctness while maintaining diversity; and (ii) syntax-aware temperature adaptation, which classifies tokens by their syntactical and functional roles and adjusts the sampling temperature accordingly, enforcing low temperature for syntax-critical tokens and higher temperature for exploratory ones. Our approach operates entirely at inference time without requiring any additional model fine-tuning. Through evaluations on multiple open-source LLMs using the VerilogEval benchmark, we demonstrate significant improvements in syntactic validity, functional correctness, and output diversity, while the execution overhead (performance overhead) is imperceptible.