This paper introduces a novel benchmark dataset designed to evaluate the capabilities of Vision Language Models (VLMs) on tasks that combine visual reasoning with subject-specific background knowledge in the German language. In contrast to widely used English-language benchmarks that often rely on artificially difficult or decontextualized problems, this dataset draws from real middle school curricula across nine domains including mathematics, history, biology, and religion. The benchmark includes over 2,000 open-ended questions grounded in 486 images, ensuring that models must integrate visual interpretation with factual reasoning rather than rely on superficial textual cues. We evaluate thirteen state-of-the-art open-weight VLMs across multiple dimensions, including domain-specific accuracy and performance on adversarial crafted questions. Our findings reveal that even the strongest models achieve less than 45% overall accuracy, with particularly poor performance in music, mathematics, and adversarial settings. Furthermore, the results indicate significant discrepancies between success on popular benchmarks and real-world multimodal understanding. We conclude that middle school-level tasks offer a meaningful and underutilized avenue for stress-testing VLMs, especially in non-English contexts. The dataset and evaluation protocol serve as a rigorous testbed to better understand and improve the visual and linguistic reasoning capabilities of future AI systems.

Wer sich mit medizinrechtlichen Fragestellungen befasst, wird unweigerlich früher oder später auf das umfassende Wirken von Dr. Rudolf Ratzel stoßen. Denn schon zu Zeiten, als es das heute so dynamische Rechtsgebiet „Medizinrecht“ nicht einmal terminologisch gab, war er der Sache nach längst in diesem Bereich anwaltlich sowie als Autor von Fachveröffentlichungen tätig. Die Festschrift versammelt Beiträge in der ganzen Bandbreite des Medizinrechts. Sie zeugt in etwa so von „Medizinrecht von A-Z“ wie Dr. Ratzel und sein berufliches Schaffen selbst. 

As universities around the world welcome increasing numbers of international students, there is a growing demand for scalable, objective tools that can support both language learning and applicant selection based on spoken language proficiency. In particular, pronunciation and comprehension remain persistent challenges for non-native speakers and are key factors for communication in academic environments. Traditional methods of assessing these skills are labor-intensive or often rely on surface-level metrics such as transcription accuracy, which do not fully capture a learner’s communicative competence. This work introduces TalkPro, a multi-modal system for pronunciation and comprehension assessment as well as language learning, designed to address this need. The system provides continuous, personalized feedback on learners’ spoken language, with a specific focus on phoneme-level accuracy as well as articulatory patterns. Instead of relying solely on conventional speech recognition outputs, which are often able to compensate even major pronunciation errors, TalkPro generates detailed acoustic analyses that pinpoint learner-specific difficulties. These include not only phoneme-level errors but also recurring articulatory tendencies, such as misplacement of the tongue, incorrect voicing, or inappropriate manner of articulation. The system also includes a text-to-speech (TTS) engine to generate spoken content adapted to vocabulary gaps, which is then followed by targeted comprehension questions. TTS can also be used to test listening comprehension either word by word in a dictation style or semantically using a large language model (LLM) as a judge. Overall, these components form an integral approach to pronunciation and comprehension training in a blended learning environment and can also be used for automated assessment. Preliminary experiments with incoming students from India to Germany indicate that phoneme-level ASR effectively identifies pronunciation errors, whereas grapheme-level ASR tends to overlook them. Future research will involve a comprehensive evaluation of automated results against human judgment, alongside the expansion of TalkPro's training capabilities with LLM-based reading comprehension modules that prioritize conceptual understanding over traditional verbatim recall.

Background

Traumatic brain injury (TBI) is one of the most important pediatric conditions worldwide. In Germany, hospitalization rates for mild TBI drastically exceed hospitalization rates from similar healthcare systems.

Methods

The SaVeBRAIN.Kids trial will implement and test a novel care pathway (nCP) for evidence-based standardized risk assessment, structured observation in the emergency department (ED) for several hours, and technology-supported home monitoring with the aim to reduce hospitalizations. This non-inferiority multicenter study will be carried out using a cluster-randomized stepped-wedge design, with all centers starting in the control phase and sequentially transitioning to the intervention. Eligible participants (age ≥ 3 months and < 18 years) must present within 48 h of head injury, have minimal symptoms (Glasgow coma scale ≥ 14), and no risk factors for intracranial complications. The co-primary outcomes are the relative risk of hospitalization and the proportion of unplanned re-visits within 72 h of presentation to the ED for ambulatory cases. Secondary outcomes include clinical safety measures, cost-effectiveness, and process evaluation. Based on power calculations (α = 0.05, power = 0.9), 1390 patients will be recruited over 12 months.

Discussion

 The SaVeBRAIN.Kids trial addresses a relevant healthcare challenge by testing a new approach to pediatric mild TBI management in Germany. It aligns with current evidence while accounting for the country’s specific healthcare context. If successful, the intervention could substantially reduce unnecessary hospitalizations and free inpatient capacities while preserving patient safety.

Trial registration

German Clinical Trials Registry (DRKS00035623). Registered on January 21, 2025.

Hydrogen and Syngas - Platform for a sustainable future, 28. - 29.Oktober 2025, Essen, Germany 

The ARM TrustZone Trusted Execution Environment (TEE) allows software to run in an isolated environment, separated from the untrusted OS. The isolation is based on the Secure Monitor (SM), software running at the most privileged hardware level, with unrestricted access to all system resources, including those of the TEE. Critically, recent research revealed widespread vulnerabilities in SMs that break the TEE isolation and, thus, undermine the very purpose of the TEE. To this end, we present SH3ARS, a fundamental restructuration of the SM firmware that reduces the privileges of the SM and restores ARM TrustZone isolation. SH3ARS modifies the SM to irrevocably relinquish access to memory outside its own address space through a page table latching mechanism. Furthermore, we introduce guards, carefully crafted, gadget-free code sequences, that supervise the context switch to and from the TEE, preventing code-reuse attacks against the TEE – a technique we refer to as SMC-oriented programming. Relying on software changes, SH3ARS ensures TEE isolation guarantees, even if the SM is compromised. We apply SH3ARS to the reference implementation of the SM on ARMv8.0-A, as deployed on millions of devices. We implement a proof of concept on real hardware, and our evaluation shows that the overhead is lower than 6% for most workloads.

Aktau Dialogue: Green Hydrogen 2025, 01.-03. Oktober 2025, Aktau, Kasachstan

An innovative training model named Medical Angiography Simulation Phantom for Physician Training (MANTA) has been developed at a university makerspace to enhance the education of future interventional radiologists. This model enables medical trainees to practice minimally invasive procedures for treating vascular occlusions in a realistic environment, without the need to operate directly on patients or animal models.

The phantom is produced using 3D printing technology and standard materials, resulting in a cost-effective solution with material expenses around €250. Its design includes a transparent, flexible vascular system integrated with a controllable pump that simulates human arterial blood pressure and flow rates, providing physiological conditions akin to real-life interventions. This setup allows for the safe rehearsal of catheter handling, guidewire manipulation, and occlusion techniques using coils.

The development of a cost-effective open source approach is intended to facilitate the use of medical technology in emerging or underdeveloped countries. By offering a practical and realistic training tool, the MANTA phantom aims to improve the proficiency of upcoming radiologists, ultimately enhancing patient care during critical vascular interventions.