Large language models (LLMs) have gained widespread adoption across diverse domains and applications. However, as LLMs become more integrated into various systems, concerns around their security are growing. Existing relevant studies mainly focus on threats arising from user prompts (e.g., prompt injection attack) and model output (e.g. We introduce system prompt poisoning, a new attack vector against LLMs that, unlike traditional user prompt injection, poisons system prompts and persistently impacts all subsequent user interactions and model responses. We propose three practical attack strategies: brute-force poisoning, adaptive in-context poisoning, and adaptive chain-of-thought (CoT) poisoning, and introduce Auto-SPP, a framework that automates the poisoning of system prompts with these strategies. Our comprehensive evaluation across four reasoning and non-reasoning LLMs, four distinct attack scenarios, and two challenging domains (mathematics and coding) reveals the attack's severe impact. The findings demonstrate that system prompt poisoning is not only highly effective, drastically degrading task performance in all scenario-strategy combinations, but also persistent and robust, remaining potent even when user prompts employ prompting-augmented techniques like CoT. Critically, our results highlight the stealthiness of this attack by showing that current black-box based prompt injection defenses cannot effectively defend against it. Large language models (LLMs) like GPT -5 (OpenAI, 2025), Gemini 2.5 (Gemini Team and Google, 2023), and Claude Opus 4.1 (Anthropic, 2025) have shown exceptional performance, driving their widespread integration into the modern software ecosystem. This includes domain-specific applications like Cursor (Anysphere, Inc., 2025) and Adobe Firefly (Adobe, 2025), development frameworks such as Langchain (Harrison Chase, 2025) and Promptflow (Microsoft, 2025), and research communities like Hugging Face (Face, 2025) and HELM (Liang et al., 2022). The proliferation of LLMs has heightened security concerns, with popular commercial platforms (e.g., ChatGPT, Gemini) exhibiting vulnerabilities such as data poisoning and jailbreaks (Zou et al., 2023a; Fu et al., 2024; Bowen et al., 2024). This risk extends across the entire LLM ecosystem, where studies show data abuse and privacy violations are are frequently reported (Hou et al., 2024; Iqbal et al., 2024; Huang et al., 2024). Prompts in LLMs are typically categorized into two types: user prompt and system prompt. User prompt refers to the input provided by the end-user that is meant to get a specific response from language model.

Unmanned Aerial Vehicles (UA Vs), commonly referred to as drones, have witnessed a remarkable surge in popularity due to their versatile applications. These cyber-physical systems depend on multiple sensor inputs, such as cameras, GPS receivers, accelerometers, and gyroscopes, with faults potentially leading to physical instability and serious safety concerns. To mitigate such risks, anomaly detection has emerged as a crucial safeguarding mechanism, capable of identifying the physical manifestations of emerging issues and allowing operators to take preemptive action at runtime. Recent anomaly detection methods based on LSTM neural networks have shown promising results, but three challenges persist: the need for models that can generalise across the diverse mission profiles of drones; the need for interpretability, enabling operators to understand the nature of detected problems; and the need for capturing domain knowledge that is difficult to infer solely from log data. Motivated by these challenges, this paper introduces RADD, an integrated approach to anomaly detection in drones that combines rule mining and unsupervised learning. In particular, we leverage rules (or invariants) to capture expected relationships between sensors and actuators during missions, and utilise unsupervised learning techniques to cover more subtle relationships that the rules may have missed. We implement this approach using the ArduPilot drone software in the Gazebo simulator, utilising 44 rules derived across the main phases of drone missions, in conjunction with an ensemble of five unsupervised learning models. We find that our integrated approach successfully detects 93.84% of anomalies over six types of faults with a low false positive rate (2.33%), and can be deployed effectively at runtime. Furthermore, RADD outperforms a state-of-the-art LSTM-based method in detecting the different types of faults evaluated in our study.

How to allocate limited resources to projects that will yield the greatest long-term benefits is a problem that often arises in decision-making under uncertainty. For example, organizations may need to evaluate and select innovation projects with risky returns. Similarly, when allocating resources to research projects, funding agencies are tasked with identifying the most promising proposals based on idiosyncratic criteria. Finally, in participatory budgeting, a local community may need to select a subset of public projects to fund. Regardless of context, agents must estimate the uncertain values of a potentially large number of projects. Developing parsimonious methods to compare these projects, and aggregating agent evaluations so that the overall benefit is maximized, are critical in assembling the best project portfolio. Unlike in standard sorting algorithms, evaluating projects on the basis of uncertain long-term benefits introduces additional complexities. We propose comparison rules based on Quicksort and the Bradley--Terry model, which connects rankings to pairwise "win" probabilities. In our model, each agent determines win probabilities of a pair of projects based on his or her specific evaluation of the projects' long-term benefit. The win probabilities are then appropriately aggregated and used to rank projects. Several of the methods we propose perform better than the two most effective aggregation methods currently available. Additionally, our methods can be combined with sampling techniques to significantly reduce the number of pairwise comparisons. We also discuss how the Bradley--Terry portfolio selection approach can be implemented in practice.

With the rapid expansion of content on social media platforms, analyzing and comprehending online discourse has become increasingly complex. This paper introduces LLMTaxo, a novel framework leveraging large language models for the automated construction of taxonomies of factual claims from social media by generating topics at multiple levels of granularity. The resulting hierarchical structure significantly reduces redundancy and improves information accessibility. We also propose dedicated taxonomy evaluation metrics to enable comprehensive assessment. Evaluations conducted on three diverse datasets demonstrate LLMTaxo's effectiveness in producing clear, coherent, and comprehensive taxonomies. Among the evaluated models, GPT-4o mini consistently outperforms others across most metrics. The framework's flexibility and low reliance on manual intervention underscore its potential for broad applicability.

Detecting AI-involved text is essential for combating misinformation, plagiarism, and academic misconduct. However, AI text generation includes diverse collaborative processes (AI-written text edited by humans, human-written text edited by AI, and AI-generated text refined by other AI), where various or even new LLMs could be involved. Texts generated through these varied processes exhibit complex characteristics, presenting significant challenges for detection. Current methods model these processes rather crudely, primarily employing binary classification (purely human vs. AI-involved) or multi-classification (treating human-AI collaboration as a new class). We observe that representations of texts generated through different processes exhibit inherent clustering relationships. Therefore, we propose DETree, a novel approach that models the relationships among different processes as a Hierarchical Affinity Tree structure, and introduces a specialized loss function that aligns text representations with this tree. To facilitate this learning, we developed RealBench, a comprehensive benchmark dataset that automatically incorporates a wide spectrum of hybrid texts produced through various human-AI collaboration processes. Our method improves performance in hybrid text detection tasks and significantly enhances robustness and generalization in out-of-distribution scenarios, particularly in few-shot learning conditions, further demonstrating the promise of training-based approaches in OOD settings. Our code and dataset are available at https://github.com/heyongxin233/DETree.

Autonomous agents rely on automated planning algorithms to achieve their objectives. Simulation-based planning offers a significant advantage over declarative models in modelling complex environments. However, relying solely on a planner that produces a single plan may not be practical, as the generated plans may not always satisfy the agent's preferences. To address this limitation, we introduce $\texttt{FBI}_\texttt{LTL}$, a diverse planner explicitly designed for simulation-based planning problems. $\texttt{FBI}_\texttt{LTL}$ utilises Linear Temporal Logic (LTL) to define semantic diversity criteria, enabling agents to specify what constitutes meaningfully different plans. By integrating these LTL-based diversity models directly into the search process, $\texttt{FBI}_\texttt{LTL}$ ensures the generation of semantically diverse plans, addressing a critical limitation of existing diverse planning approaches that may produce syntactically different but semantically identical solutions. Extensive evaluations on various benchmarks consistently demonstrate that $\texttt{FBI}_\texttt{LTL}$ generates more diverse plans compared to a baseline approach. This work establishes the feasibility of semantically-guided diverse planning in simulation-based environments, paving the way for innovative approaches in realistic, non-symbolic domains where traditional model-based approaches fail.

Large language models (LLMs) are transforming education by answering questions, explaining complex concepts, and generating content across a wide range of subjects. Despite strong performance on academic benchmarks, they often fail to tailor responses to students' grade levels. This is a critical need in K-12 education, where age-appropriate vocabulary and explanation are essential for effective learning. Existing models frequently produce outputs that are too advanced or vague for younger learners, and there are no standardized benchmarks to evaluate their ability to adjust across cognitive and developmental stages. To address this gap, we introduce EduAdapt, a benchmark of nearly 48k grade-labeled QA pairs across nine science subjects, spanning Grades 1-12 and grouped into four grade levels. We evaluate a diverse set of open-source LLMs on EduAdapt and find that while larger models generally perform better, they still struggle with generating suitable responses for early-grade students (Grades 1-5). Our work presents the first dataset and evaluation framework for assessing grade-level adaptability in LLMs, aiming to foster more developmentally aligned educational AI systems through better training and prompting strategies. EduAdapt code and datasets are publicly available at https://github.com/NaumanNaeem/EduAdapt.

Optimizing the energy management within a smart grids scenario presents significant challenges, primarily due to the complexity of real-world systems and the intricate interactions among various components. Reinforcement Learning (RL) is gaining prominence as a solution for addressing the challenges of Optimal Power Flow (OPF) in smart grids. However, RL needs to iterate compulsively throughout a given environment to obtain the optimal policy. This means obtaining samples from a, most likely, costly simulator, which can lead to a sample e fficiency problem. In this work, we address this problem by substituting costly smart grid simulators with surrogate models built using Physics-Informed Neural Networks (PINN)s, optimizing the RL policy training process by arriving to convergent results in a fraction of the time employed by the original environment. Specifically, we tested the performance of our PINN surrogate against other state-of-the-art data-driven surrogates and found that the understanding of the underlying physical nature of the problem makes the PINN surrogate the only method that we studied capable of learning a good RL policy, in addition to not having to use samples from the real simulator. Our work shows that, by employing PINN surrogates, we can improve training speed by 50%, comparing to training the RL policy by not using any surrogate model, enabling us to achieve results with score on par with the original simulator more rapidly. Keywords: Smart Grids Control, Reinforcement Learning, Physics-informed Neural Networks, Active Network Management, Optimal Power Flow, Surrogate Models, Renewable EnergyRL Reinforcement Learning EA Expert agent PINN Physics-Informed Neural Networks ANN Artificial Neural Network OPF Optimal Power Flow ESS Energy Storage Systems SoC State of Change MAE Mean Absolute Error 1. Introduction Smart grids are a pivotal concept driving the current modernization of electrical networks, addressing the urgent need to reduce greenhouse gas emissions, enhance energy e fficiency, and improve grid stability through demand response mechanisms. The European Union aims to achieve 43% renewable energy generation by 2030 [1], and in 2021, the renewable energy share rose to 32 .1% [2]. Corresponding author Email address: julen.cestero@polimi.it Modern societies require advanced grids capable of predicting and mitigating the uncertainties associated with renewable energy sources.

Throughout application domains, we now rely extensively on algorithmic systems to engage with ever-expanding datasets of information. Despite their benefits, these systems are often complex (comprising of many intricate tools, e.g., moderation, recommender systems, prediction models), of unknown structure (due to the lack of accompanying documentation), and having hard-to-predict yet potentially severe downstream consequences (due to the extensive use, systematic enactment of existing errors, and many comprising feedback loops). As such, understanding and evaluating these systems as a whole remains a challenge for both researchers and legislators. To aid ongoing efforts, we introduce a formal framework for such visibility allocation systems (VASs) which we define as (semi-)automated systems deciding which (processed) data to present a human user with. We review typical tools comprising VASs and define the associated computational problems they solve. By doing so, VASs can be decomposed into sub-processes and illustrated via data flow diagrams. Moreover, we survey metrics for evaluating VASs throughout the pipeline, thus aiding system diagnostics. Using forecasting-based recommendations in school choice as a case study, we demonstrate how our framework can support VAS evaluation. We also discuss how our framework can support ongoing AI-legislative efforts to locate obligations, quantify systemic risks, and enable adaptive compliance.

The great rivers of Bangladesh, arteries of commerce and sustenance, are also agents of relentless destruction. Each year, they swallow whole villages and vast tracts of farmland, erasing communities from the map and displacing thousands of families. To track this slow-motion catastrophe has, until now, been a Herculean task for human analysts. Here we show how a powerful general-purpose vision model, the Segment Anything Model (SAM), can be adapted to this task with remarkable precision. To do this, we assembled a new dataset - a digital chronicle of loss compiled from historical Google Earth imagery of Bangladesh's most vulnerable regions, including Mokterer Char Union, Kedarpur Union, Balchipara village, and Chowhali Upazila, from 2003 to 2025. Crucially, this dataset is the first to include manually annotated data on the settlements that have vanished beneath the water. Our method first uses a simple color-channel analysis to provide a rough segmentation of land and water, and then fine-tunes SAM's mask decoder to recognize the subtle signatures of riverbank erosion. The resulting model demonstrates a keen eye for this destructive process, achieving a mean Intersection over Union of 86.30% and a Dice score of 92.60% - a performance that significantly surpasses traditional methods and off-the-shelf deep learning models. This work delivers three key contributions: the first annotated dataset of disappeared settlements in Bangladesh due to river erosion; a specialized AI model fine-tuned for this critical task; and a method for quantifying land loss with compelling visual evidence. Together, these tools provide a powerful new lens through which policymakers and disaster management agencies can monitor erosion, anticipate its trajectory, and ultimately protect the vulnerable communities in its path.