Il s'agit de notre dernière sélection de publications et de brevets mondiaux en anglais sur l'ingénierie des prompts, parmi de nombreuses revues scientifiques en ligne, classées et axées sur l'ingénierie des prompts, la conception des prompts, la contextualisation, la variabilité des prompts, l'ajustement des hyperparamètres, l'atténuation des biais, l'apprentissage à zéro coup et l'apprentissage à quelques coups.
Input/output-component (io-component) for a control device for controlling a device or system
Patent published on the 2026-04-15 in EP under Ref EP4726483 by SIEMENS AG [DE] (Haneder Thomas [de], Lindemann Lars [de], Rost Julian [de])
Abstract: [0001] The present invention describes an input/output-component (IO-component) for a control device for controlling a device or system, wherein the IO-component is designed and set up - for input and/or output of IO control data into and/or out of the control device via a communication link using a communication protocol, - and for connection to a middleware component of the control device, and wherein the IO-component is designed and set up for converting IO control data received via the commu[...]
Our summary: The invention describes an IO-component for a control device that manages input and output of control data. It converts IO control data into middleware variables independent of the communication protocol. The conversion process includes normalization, standardization, logical structuring, and contextualization of the data.
IO-component, control device, communication protocol, middleware variables
Patent
A Method for 3D Building Individualization Integrating SAMPolyBuild and Multiple Spatial-Geometric Features
Published on 2026-02-03 by Lianshuai Cao, Yi Cheng, Zheng Zhang, Ge Zhu, Kunyang Ma, Xinyue Xu @MDPI
Abstract: Individualization of buildings is one of the key issues in the establishment of three-dimensional (3D) building models. Most existing individualization methods rely on inefficient manual separation, while deep learning approaches require extensive pre-training and are highly influenced by the spatial structure of the models. To address these issues, this paper proposes a novel method for 3D building individualization that integrates SAMPolyBuild with multiple spatial-geometric features. Leveragi[...]
Our summary: This paper presents a method for 3D building individualization that integrates SAMPolyBuild with spatial-geometric features. The approach utilizes zero-shot learning for initial extraction and refines accuracy through statistical parameters. Experiments show effective extraction of building models, achieving an F1-score of about 0.83.
3D Building Individualization, SAMPolyBuild, Spatial-Geometric Features, Zero-Shot Learning
Publication
Deep Reinforcement Learning-Driven Adaptive Prompting for Robust Medical LLM Evaluation
Published on 2026-02-02 by Dong Ding, Wang Xi, Zenghui Ding, Jianqing Gao @MDPI
Abstract: The accurate and reliable evaluation of large language models (LLMs) in medical domains is critical for real-world clinical deployment, automated medical reasoning, and patient safety. However, the evaluation process is highly sensitive to prompt design, and prevalent reliance on fixed or randomly sampled prompt policies often fails to dynamically adapt to clinical context, question complexity, or evolving safety requirements. This article presents a novel reinforcement learning-based framework [...]
Our summary: This article introduces a reinforcement learning-based framework for adaptive prompt selection in medical LLM evaluation. It formulates prompt selection as a Markov Decision Process and utilizes a deep Q-Network to enhance evaluation metrics. Experiments show consistent improvements in accuracy, safety, and medical terminology coverage across multiple datasets.
Reinforcement Learning, Prompt Optimization, Medical LLM, Evaluation Framework
Publication
Weak-to-Strong Honesty Alignment via Group-Relative Policy Optimization
Published on 2026-01-30 by Jie Zhang, Yunfan Xie, Wen Zou @MDPI
Abstract: Ensuring that Large Language Models align with human values of honesty is a critical challenge, particularly due to the scarcity of labeled data for distinguishing known versus unknown knowledge boundaries. We propose a weak-to-strong generalization framework utilizing Group Relative Policy Optimization (GRPO). Unlike standard supervised fine-tuning or prompt engineering, our framework trains a lightweight “honest head” to rank response candidates based on multifacete[...]
Our summary: We propose a weak-to-strong generalization framework for aligning large language models with human honesty values. Our method utilizes Group Relative Policy Optimization to train an "honest head" that ranks response candidates based on honesty scores. Experiments show significant performance improvements over existing methods in honesty alignment across various datasets.
Honesty Alignment, Group Relative Policy Optimization, Large Language Models, Self-labeling
Publication
Machine Learning Models for Reliable Gait Phase Detection Using Lower-Limb Wearable Sensor Data
Published on 2026-01-29 by Muhammad Fiaz, Rosita Guido, Domenico Conforti @MDPI
Abstract: Accurate gait-phase detection is essential for rehabilitation monitoring, prosthetic control, and human–robot interaction. Artificial intelligence supports continuous, personalized mobility assessment by extracting clinically meaningful patterns from wearable sensors. A richer view of gait dynamics can be achieved by integrating additional signals, including inertial, plantar flex, footswitch, and EMG data, leading to more accurate and informative gait analysis. Motivated by these [...]
Our summary: This study focuses on accurate gait-phase detection using lower-limb wearable sensor data. It employs machine learning models, including tree-based ensembles, to classify gait phases with over 97% accuracy. The research highlights the importance of integrating multiple sensor modalities for improved gait analysis.
Gait Phase Detection, Machine Learning, Wearable Sensors, Sensor Fusion
Publication
Hierarchical Prompt Engineering and Adaptive Evaluation for Reliable Synthetic Knowledge Dialogues
Published on 2026-01-29 by Hyeongju Ju, EunKyeong Lee, Junyoung Kang, JaKyoung Kim, Dongsuk Oh @MDPI
Abstract: Large Language Models (LLMs) have demonstrated exceptional performance in knowledge-based dialogue generation and text evaluation. Synthetic data serves as a cost-effective alternative for generating high-quality datasets. However, it often plagued by hallucinations, inconsistencies, and self-anthropomorphized responses. Concurrently, manual construction of knowledge-based dialogue datasets remains bottlenecked by prohibitive costs and inherent human subjectivity. To address these multifaceted c[...]
Our summary: The paper presents ACE, a hybrid method for constructing high-quality knowledge-based dialogue datasets using hierarchical prompt engineering. It addresses issues of hallucinations and response consistency while introducing the Truthful Answer Score for improved evaluation. Experimental results show ACE outperforms existing benchmarks and enhances the reliability of synthetic datasets.
Hierarchical Prompt Engineering, Adaptive Evaluation, Synthetic Knowledge Dialogues, Truthful Answer Score
Publication
Comparative Analysis and Optimisation of Machine Learning Models for Regression and Classification on Structured Tabular Datasets
Published on 2026-01-29 by Siegfried Fredrich Stumpfe, Sandile Charles Shongwe @MDPI
Abstract: This research entails comparative analysis and optimisation of machine learning models for regression and classification tasks on structured tabular datasets. The primary target audience for this analysis comprises researchers and practitioners working with structured tabular data. Common fields include biostatistics, insurance, and financial risk modelling, where computational efficiency and robust predictive performance are essential. Four machine learning techniques (i.e., linear/logistic reg[...]
Our summary: This research conducts a comparative analysis and optimization of machine learning models for regression and classification on structured tabular datasets. It evaluates four techniques across 72 datasets, revealing that dataset characteristics significantly influence model performance. The findings emphasize the robustness of linear models and the superior performance of non-linear models in complex environments.
Machine Learning, Regression, Classification, Tabular Datasets
Publication
Detection of site phishing using neural network-enabled site image analysis leveraging few-shot learning
Patent published on the 2025-09-18 in US under Ref US2025294055 by AKAMAI TECH INC [US] (Costa Nadav George [il])
Abstract: Website phishing detection is enabled using a siamese neural network. One twin receives a query image associated with a website page. The other twin receives a subset of a set of reference website images together with positive (phishing) examples that were used to train the networks, the subset of reference website images having been determined by applying an identifier associated with a brand of interest. The operation of applying the identifier significantly reduces the relevant search space f[...]
Our summary: This study presents a method for phishing detection using a siamese neural network that analyzes website images. One network twin processes a query image, while the other processes a subset of reference images. The approach reduces the search space for identifying phishing sites and triggers mitigation actions based on inference results.
phishing detection, neural networks, few-shot learning, image analysis
Patent
