#01Aug 3, 2026
cs.CL
CTRAG: An In-Context Retrieval-based Framework for Automated Compliance Checking using LLMs
Muhammad Roman, Karen Rafferty, Barry Devereux
Trust is fundamental in modern regulatory ecosystems, and compliance checking plays a critical role in fostering that trust. Regulatory compliance verification is essential for businesses operating in highly controlled environments, as it ensures alignment with sector-specific guidelines across domains such as financial reporting, data privacy, and cybersecurity. Manual compliance testing, however, is often time-intensive and prone to inconsistencies, particularly when compliance depends indirectly on third-party services such as cloud providers, where vendors rely on external providers to meet regulatory standards. In this paper, we present CTRAG, a novel Retrieval-Augmented Generation (RAG) pipeline designed for automated compliance checking. CTRAG employs advanced strategies, including adaptive chunking, dynamic retrieval configurations, and in-context learning, to improve the precision and relevance of compliance assessments. By extracting control questions from regulatory texts and cross-referencing them with unstructured company documentation, CTRAG achieves highly accurate, document-informed compliance verification, even in cases of indirect compliance through third-party services. Empirical evaluations demonstrate significant improvements, with CTRAG achieving an F1-score of 78% and a recall of 85% in the final deployed configuration, ensuring minimal missed non-compliance cases while reducing manual reviewer effort in a real-world deployment. To validate CTRAG value, we developed and deployed a POC within a Big Four professional services firm, applying it to real-world cases and cross-checking results against manual compliance reports. These findings highlight CTRAG potential to streamline compliance workflows, mitigate risks, and enhance regulatory trust in complex, high-stakes environments.
#02Aug 3, 2026
cs.AI
Magnet: Detecting Cross-Session AI Misuse Through Capability Accumulation
Natalie Isak, Matthew Dressman
The most capable AI deployments are not single models but ensembles of specialized agents that delegate and act in coordination. This architecture unlocks powerful new capabilities, and it also introduces risks that existing frameworks for monitoring, detection, and mitigation were not designed to address. Most state-of-the-art AI abuse detection literature focuses on single-turn or multi-turn (single-session) threat models. This leaves a critical gap: an attacker can decompose a harmful goal into innocuous-looking units and execute each in isolated agentic sessions. The agent is stateless between conversations, but the attacker is not. This asymmetry allows for cross-session trajectories that are effective at evading detection. Our contributions are twofold. First, we demonstrate cross-session goal decomposition as an evasion technique, showing it may elicit more harmful capability than equivalent single-session or multi-turn attacks. By capability we mean an artifact produced at one step of an objective, evidenced by what an interaction produced (model responses and tool-call results), and composable with capabilities accrued elsewhere into a harmful whole. Second, we propose Magnet: an efficient and robust detection approach that models relevant capabilities accrued over time and across agentic conversations, aggregated at a higher-level correlator (in this case, a user ID) rather than per-conversation state. The main challenge is assembling the evidence bundle Magnet reasons over. The incriminating artifacts may be needles scattered through a haystack of benign sessions that are individually harmless, dangerous only once collected. Rather than searching the haystack straw-by-straw (i.e. per-session inspection), Magnet does what its name implies: it attracts the relevant needles out of the hay, across sessions and across time, into a compact evidence bundle a detector can act on.
#03Aug 3, 2026
eess.SY
Bridging Artificial Intelligence and Power Systems Education Using a Hands-On Executable Framework
Junjie Yin, Buxin She, Xinyu Feng and 2 more
Artificial intelligence (AI) is increasingly central to power and energy systems, supporting modeling, forecasting, optimization, and control. Yet most existing works emphasize specialized applications and offer little reusable material for newcomers or interdisciplinary learners, who increasingly rely on large language models rather than building their own. This gap points to a need for engineering-grounded AI (EGAI), in which AI workflows follow established engineering and power-system domain rules rather than acting as task-agnostic black boxes. Motivated by a community survey of researchers and practitioners, which shows 92% report at least one barrier before running an AI model and 94% want a power-specific hands-on course. This paper presents a framework consisting of open, executable module library that lowers the entry barrier for AI in power systems. The modules follow a progressive difficulty ladder that maps core AI concepts onto representative power-system tasks: (i) foundational deep neural network (DNN) templates for function approximation and load-curve fitting; (ii) a domain-coupled convolutional neural network (CNN) power-flow surrogate for a 5-bus system; and (iii) frontier modules on DNN-assisted optimization, deep reinforcement learning (DRL) for battery storage control, and physics-informed neural networks (PINNs) for the swing equation. All modules are released as Jupyter notebooks that run locally or on Google Colab and are delivered through an IEEE online course and IEEE Power & Energy Society (PES) webinar series. The webinar drew more than 590 live attendees, which is among the ten most-attended IEEE PES webinars, and over 344 repository visits within two weeks, reinforcing the survey-based motivation.
#04Aug 3, 2026
cs.CL
MedPRESS: A Multi-turn Benchmark for Patient-Pressure-Induced Medical Sycophancy in LLMs
Saman Sarker Joy, Niloy Farhan
Large language models (LLMs) are increasingly used for health-related advice. Existing research measures their safety with static questions rather than pressured patient-facing conversations. We introduce MedPRESS, a multi-turn benchmark for measuring patient-pressure-induced sycophancy in LLMs. MedPRESS contains 600 medically grounded five-turn dialogues across three scenario families: medication and treatment demand, personal health self-care, and symptom triage and care resistance. Each dialogue begins with a health query and escalates through personal experience, social proof, external evidence claims, and direct adversarial challenge. We evaluate 20 LLMs across general, medical-domain, lightweight, large, open-weight, and proprietary families using structured judging and safety-focused metrics. Results show that models frequently shift toward unsafe agreement under repeated patient pressure, with substantial variation across model families, model scale, and prompt type. Anti-sycophancy prompting improves robustness for several models, but does not eliminate unsafe agreement. MedPRESS highlights a critical gap in medical LLM evaluation: safe medical knowledge is not enough unless models can maintain it under conversational pressure.
#05Aug 3, 2026
cs.SE
SWE-Touch: Benchmarking Coding Agents When Users Touch the Code
Yuqiao Tan, Jinxiang Meng, Fangyu Lei and 4 more
Real-world software development requires coding agents to operate in shared workspaces where users may inspect and modify code during an ongoing task, yet existing repository-level benchmarks typically evaluate agents working alone or restrict user participation to messages. This leads us to ask: how do coding agents understand and respond to code changes in a shared workspace? We introduce SWE-Touch, a framework that stress-tests this setting through validated Counter-Edits: plausible edits to task-relevant code that conflict with task completion. SWE-Touch mines task-critical regions from multiple repair trajectories, uses a separate User Patch Generator to construct the edits, and injects them with contextual user messages when agents reach the relevant code. We evaluate nine coding models on SWE-bench Verified, with additional experiments on longer-horizon tasks from SWE-Bench Pro and DeepSWE. Counter-Edit lowers average resolve rate by 7.7 percentage points on SWE-bench Verified, with degradation also persisting on both longer-horizon benchmarks. Trajectory analysis links these failures to limited awareness of the evolving workspace: agents may retain conflicting code or replace it without sufficiently re-inspecting the repository and validating the revised code with targeted tests. These findings show that strong autonomous performance does not yet ensure the state awareness and adaptive behavior needed for shared-workspace collaboration, and point to detecting workspace changes, reconciling conflicting edits with the task, and verifying the affected behavior as key capabilities for future optimization.