A schema‐first alignment framework builds compact, executable domain‐specific language models under data scarcity. Large‐scale synthetic question‐answer generation instills domain knowledge, and a code‐centric IR‐to‐DPO pipeline aligns generation with tool‐executable syntax.
Di Wang +4 more
wiley +1 more source
The Severity of Phonological Deficit as a Determinant of Dyslexia Compensation in Adulthood. [PDF]
Pellicer-Magraner A +3 more
europepmc +1 more source
Exploiting Edge Semantics in Job Shop Scheduling Problem With Heterogeneous Graph Transformers
A heterogeneous graph transformer (HGT) is introduced for reinforcement learning‐based job shop scheduling by explicitly distinguishing precedence and machine‐contention relations through edge‐type‐specific attention. The proposed framework learns richer scheduling representations, improves decision quality over homogeneous graph models, and highlights
Bulent Soykan, Fatih Kasimoglu
wiley +1 more source
A natural language processing approach to determine whether streamlined health technology assessment for multi-indication medicines is feasible. [PDF]
Luc I +4 more
europepmc +1 more source
An large language model‐powered multimodal framework is developed for robotic endoscope control. It integrates speech recognition and real‐time instrument tracking, achieving 89.47% command accuracy with ~1s latency for natural human–robot interaction in minimally invasive surgery.
Yisen Huang +7 more
wiley +1 more source
Localised Written-Form Deviations in MMSE-J Sentence-Writing Responses Receiving a Score of 1/1: A Retrospective Descriptive Case Series. [PDF]
Kowari T, Shinonaga A, Ikeno M.
europepmc +1 more source
Revealing Hidden Raman Signatures Through Attention‐Based Spectral Unmixing
Weak Raman signatures are recovered from background‐dominated spectra using a transformer‐based, reference‐free spectral unmixing AI framework. Self‐attention reconstructs substrate contribution directly from mixed data, enabling reliable extraction of previously inaccessible vibrational features.
Dmitriy A. Poteryayev +9 more
wiley +1 more source
Sino-US-DrugQA: A benchmark for evaluating large language models in cross-jurisdictional pharmaceutical regulation. [PDF]
Fu X, Chen Z, Lu W.
europepmc +1 more source
Secure Fusion‐X harmonizes unstructured NVD descriptions with structured CVSS/CWE/CPE metadata via decision‐level fusion, overcoming the fragility of traditional unimodal models. Automated assessment of software vulnerability exploitability is essential for intelligent cyber defense, yet its effectiveness is often hindered by unstable, delayed, or ...
Mona Dolati +3 more
wiley +1 more source
Training-Free Skeleton-Semantic Keyframe Extraction for Fixed-View Industrial Assembly Video: Method Design and Case Study Evaluation. [PDF]
Li Q, Xiang H, Wang T, Zhan W.
europepmc +1 more source

