Are LLMs The W ay Forward? Abstract--Autonomous vehicle navigation in complex environments such as dense and fast-moving highways and merging scenarios remains an active area of research. In the past decade, many planning and control approaches have used reinforcement learning (RL) with notable success. However, a key limitation of RL is its reliance on well-specified reward functions, which often fail to capture the full semantic and social complexity of diverse, out-of-distribution situations. As a result, a rapidly growing line of research explores using Large Language Models (LLMs) to replace or supplement RL for direct planning and control, on account of their ability to reason about rich semantic context. However, LLMs present significant drawbacks: they can be unstable in zero-shot safety-critical settings, produce inconsistent outputs, and often depend on expensive API calls with network latency. This motivates our investigation into whether small, locally deployed LLMs ( 14B parameters) can meaningfully support autonomous highway driving through reward shaping rather than direct control. These models are attractive for practical deployment as they can run on a single GPU and avoid external API dependencies. We present a case study comparing RL-only, LLM-only, and hybrid approaches, where LLMs augment RL rewards by scoring state-action transitions during training, while standard RL policies execute at test time.

Large language models (LLMs) achieve superhuman performance on complex reasoning tasks, yet often fail on much simpler problems, raising concerns about their reliability and interpretability. We investigate this paradox through a focused study with two key design features: simplicity, to expose basic failure modes, and scale, to enable comprehensive controlled experiments. We focus on set membership queries -- among the most fundamental forms of reasoning -- using tasks like ``Is apple an element of the set \{pear, plum, apple, raspberry\}?''. We conduct a systematic empirical evaluation across prompt phrasing, semantic structure, element ordering, and model choice. Our large-scale analysis reveals that LLM performance on this elementary task is consistently brittle, and unpredictable across all dimensions, suggesting that the models' ``understanding'' of the set concept is fragmented and convoluted at best. Our work demonstrates that the large-scale experiments enabled by the simplicity of the problem allow us to map and analyze the failure modes comprehensively, making this approach a valuable methodology for LLM evaluation in general.

Automated red teaming frameworks for Large Language Models (LLMs) have become increasingly sophisticated, yet they share a fundamental limitation: their jailbreak logic is confined to selecting, combining, or refining pre-existing attack strategies. This binds their creativity and leaves them unable to autonomously invent entirely new attack mechanisms. To overcome this gap, we introduce \textbf{EvoSynth}, an autonomous framework that shifts the paradigm from attack planning to the evolutionary synthesis of jailbreak methods. Instead of refining prompts, EvoSynth employs a multi-agent system to autonomously engineer, evolve, and execute novel, code-based attack algorithms. Crucially, it features a code-level self-correction loop, allowing it to iteratively rewrite its own attack logic in response to failure. Through extensive experiments, we demonstrate that EvoSynth not only establishes a new state-of-the-art by achieving an 85.5\% Attack Success Rate (ASR) against highly robust models like Claude-Sonnet-4.5, but also generates attacks that are significantly more diverse than those from existing methods. We release our framework to facilitate future research in this new direction of evolutionary synthesis of jailbreak methods. Code is available at: https://github.com/dongdongunique/EvoSynth.

Vision-language models (VLMs) enable open-ended visual question answering but remain prone to hallucinations. We present HEDGE, a unified framework for hallucination detection that combines controlled visual perturbations, semantic clustering, and robust uncertainty metrics. HEDGE integrates sampling, distortion synthesis, clustering (entailment- and embedding-based), and metric computation into a reproducible pipeline applicable across multimodal architectures. Evaluations on VQA-RAD and KvasirVQA-x1 with three representative VLMs (LLaVA-Med, Med-Gemma, Qwen2.5-VL) reveal clear architecture- and prompt-dependent trends. Hallucination detectability is highest for unified-fusion models with dense visual tokenization (Qwen2.5-VL) and lowest for architectures with restricted tokenization (Med-Gemma). Embedding-based clustering often yields stronger separation when applied directly to the generated answers, whereas NLI-based clustering remains advantageous for LLaVA-Med and for longer, sentence-level responses. Across configurations, the VASE metric consistently provides the most robust hallucination signal, especially when paired with embedding clustering and a moderate sampling budget (n ~ 10-15). Prompt design also matters: concise, label-style outputs offer clearer semantic structure than syntactically constrained one-sentence responses. By framing hallucination detection as a geometric robustness problem shaped jointly by sampling scale, prompt structure, model architecture, and clustering strategy, HEDGE provides a principled, compute-aware foundation for evaluating multimodal reliability. The hedge-bench PyPI library enables reproducible and extensible benchmarking, with full code and experimental resources available at https://github.com/Simula/HEDGE .

Deploying embodied agents that can answer questions about their surroundings in realistic real-world settings remains difficult, partly due to the scarcity of benchmarks that faithfully capture practical operating conditions. We propose infrastructure inspection as a compelling domain for open-vocabulary Embodied Question Answering (EQA): it naturally demands multi-scale reasoning, long-range spatial understanding, and complex semantic relationships, while offering unique evaluation advantages via standardized National Bridge Inventory (NBI) condition ratings (0-9), professional inspection reports, and egocentric imagery. We introduce BridgeEQA, a benchmark of 2,200 open-vocabulary question-answer pairs (in the style of OpenEQA) grounded in professional inspection reports across 200 real-world bridge scenes with 47.93 images on average per scene. Questions require synthesizing visual evidence across multiple images and aligning responses with NBI condition ratings. We further propose a new EQA metric Image Citation Relevance to evaluate the ability of a model to cite relevant images. Evaluations of state-of-the-art vision-language models reveal substantial performance gaps under episodic memory EQA settings. To address this, we propose Embodied Memory Visual Reasoning (EMVR), which formulates inspection as sequential navigation over an image-based scene graph: images are nodes, and an agent takes actions to traverse views, compare evidence, and reason within a Markov decision process. EMVR shows strong performance over the baselines. We publicly release both the dataset and code.

Assurance for artificial intelligence (AI) systems remains fragmented across software supply-chain security, adversarial machine learning, and governance documentation. Existing transparency mechanisms - including Model Cards, Datasheets, and Software Bills of Materials (SBOMs) - advance provenance reporting but rarely provide verifiable, machine-readable evidence of model security. This paper introduces the AI Risk Scanning (AIRS) Framework, a threat-model-based, evidence-generating framework designed to operationalize AI assurance. The AIRS Framework evolved through three progressive pilot studies - Smurf (AIBOM schema design), OPAL (operational validation), and Pilot C (AIRS) - that reframed AI documentation from descriptive disclosure toward measurable, evidence-bound verification. The framework aligns its assurance fields to the MITRE ATLAS adversarial ML taxonomy and automatically produces structured artifacts capturing model integrity, packaging and serialization safety, structural adapters, and runtime behaviors. Currently, the AIRS Framework is scoped to provide model-level assurances for LLMs, but it could be expanded to include other modalities and cover system-level threats (e.g. application-layer abuses, tool-calling). A proof-of-concept on a quantized GPT-OSS-20B model demonstrates enforcement of safe loader policies, per-shard hash verification, and contamination and backdoor probes executed under controlled runtime conditions. Comparative analysis with SBOM standards of SPDX 3.0 and CycloneDX 1.6 reveals alignment on identity and evaluation metadata, but identifies critical gaps in representing AI-specific assurance fields. The AIRS Framework thus extends SBOM practice to the AI domain by coupling threat modeling with automated, auditable evidence generation, providing a principled foundation for standardized, trustworthy, and machine-verifiable AI risk documentation.

Aligning Large Language Models (LLMs) to be faithful to new knowledge in complex, multi-hop reasoning tasks is a critical, yet unsolved, challenge. We find that SFT-based methods, e.g., Reason-KE, while state-of-the-art, suffer from a "faithfulness gap": they optimize for format mimicry rather than sound reasoning. This gap enables the LLM's powerful parametric priors to override new contextual facts, resulting in critical factual hallucinations (e.g., incorrectly reasoning "Houston" from "NASA" despite an explicit edit). To solve this core LLM alignment problem, we propose Reason-KE++, an SFT+RL framework that instills process-level faithfulness. Its core is a Stage-aware Reward mechanism that provides dense supervision for intermediate reasoning steps (e.g., Decomposition, Sub-answer Correctness). Crucially, we identify that naive outcome-only RL is a deceptive trap for LLM alignment: it collapses reasoning integrity (e.g., 19.00% Hop acc) while superficially boosting final accuracy. Our process-aware framework sets a new SOTA of 95.48% on MQUAKE-CF-3k (+5.28%), demonstrating that for complex tasks, aligning the reasoning process is essential for building trustworthy LLMs.

Context: Large language models (LLMs) are released faster than users' ability to evaluate them rigorously. When LLMs underpin research, such as identifying relevant literature for systematic reviews (SRs), robust empirical assessment is essential. Objective: We identify and discuss key challenges in assessing LLM performance for selecting relevant literature, identify good (evaluation) practices, and propose recommendations. Method: Using a recent large-scale study as an example, we identify problems with the use of traditional metrics for assessing the performance of Gen-AI tools for identifying relevant literature in SRs. We analyzed 27 additional papers investigating this issue, extracted the performance metrics, and found both good practices and widespread problems, especially with the use and reporting of performance metrics for SR screening. Results: Major weaknesses included: i) a failure to use metrics that are robust to imbalanced data and do not directly indicate whether results are better than chance, e.g., the use of Accuracy, ii) a failure to consider the impact of lost evidence when making claims concerning workload savings, and iii) pervasive failure to report the full confusion matrix (or performance metrics from which it can be reconstructed) which is essential for future meta-analyses. On the positive side, we extract good (evaluation) practices on which our recommendations for researchers and practitioners, as well as policymakers, are built. Conclusions: SR screening evaluations should prioritize lost evidence/recall alongside chance-anchored and cost-sensitive Weighted MCC (WMCC) metric, report complete confusion matrices, treat unclassifiable outputs as referred-back positives for assessment, adopt leakage-aware designs with non-LLM baselines and open artifacts, and ground conclusions in cost-benefit analysis where FNs carry higher penalties than FPs.

Notice to Air Missions (NOTAMs) serve as a critical channel for disseminating key flight safety information, yet their complex linguistic structures and implicit reasoning pose significant challenges for automated parsing. Existing research mainly focuses on surface-level tasks such as classification and named entity recognition, lacking deep semantic understanding. To address this gap, we propose NOTAM semantic parsing, a task emphasizing semantic inference and the integration of aviation domain knowledge to produce structured, inference-rich outputs. To support this task, we construct Knots (Knowledge and NOTAM Semantics), a high-quality dataset of 12,347 expert-annotated NOTAMs covering 194 Flight Information Regions, enhanced through a multi-agent collaborative framework for comprehensive field discovery. We systematically evaluate a wide range of prompt-engineering strategies and model-adaptation techniques, achieving substantial improvements in aviation text understanding and processing. Our experimental results demonstrate the effectiveness of the proposed approach and offer valuable insights for automated NOTAM analysis systems. Our code is available at: https://github.com/Estrellajer/Knots.

ABSTRACT Large language models (LLMs) have shown strong reasoning capabilities and are increasingly explored for financial trading. Existing LLM-based trading agents, however, largely focus on single-step prediction and lack integrated mechanisms for risk management, which reduces their effectiveness in volatile markets. We introduce FinRS, a risk-sensitive trading framework that combines hierarchical market analysis, dual-decision agents, and multi-timescale reward reflection to align trading actions with both return objectives and downside risk constraints. Experiments on multiple stocks and market conditions show that FinRS achieves superior profitability and stability compared to state-of-the-art methods. Index T erms-- Trading Agent, Risk-Sensitive, Real Markets 1. INTRODUCTION In recent years, large language models (LLMs) [1, 2] have demonstrated significant potential in financial trading.