Product Design, Manufacturing & Innovation Resources

Dernières publications et brevets sur les grands modèles linguistiques (LLM)

Modèles de langage à grande échelle (LLM)

Astuce : suite à la sélection ci-dessous, vous pouvez rechercher et filtrer nos 2 bases de données complètes :

> outil de recherche de publications gratuit < par auteur, sujet, mots-clés, date ou revue.

> outil gratuit de recherche de brevets < pour les brevets en anglais de l'Office européen des brevets.

Il s'agit de notre dernière sélection de publications et de brevets mondiaux en anglais sur les grands modèles de langage (LLM), parmi de nombreuses revues scientifiques en ligne, classées et axées sur les thèmes suivants : grand modèle de langage, LLM, transformateur génératif pré-entraîné, pré-entraînement, architecture du transformateur, descente en gradient, GPT, tokenisation, modèle génératif, mécanisme d'auto-attention, modèle de langage masqué et MLM.

Mapping the movie-watching brain with AI-derived semantics

Published on 2026-07-10 by @MIT

Abstract: AbstractNaturalistic paradigms offer a powerful tool to investigate human brain function, but it remains difficult to link rich, continuous movie content to distributed brain activity in an interpretable way. In this study, I use a multimodal large language model (Gemini) as an automated “semantic annotator” to bridge naturalistic movie stimuli, brain responses, and cognitive performance. Using the Human Connectome Project movie-watching dataset, I segmented the film into 293 overlapping cli[...]


Our summary: This study uses a multimodal large language model to link movie content with brain responses. It segments films into clips and analyzes BOLD activation patterns in cortical regions. The findings reveal that AI-derived features can predict brain responses and relate to individual cognitive abilities.

AI, brain activity, semantic annotation, cognitive performance

Publication

Machine learning model security at a processor

Patent published on the 2026-07-02 in US under Ref US20260187254 by ADVANCED MICRO DEVICES INC [US] (Nadarajah Kathirkamanathan [ca], Blinzer Paul [us], Sanghai Kaushal Amolak [us], Subramaniam Akila [us])

Abstract: A processor protects a machine learning model (MLM) from unauthorized access. The processor employs a neural processing unit (NPU) to execute the MLM and implements decryption and encryption processes to decrypt the MLM and re-encrypt the MLM at different points along MLM storage and execution paths. Furthermore, the processor executes the encryption and decryption processes at different processing units and processing engines, thereby reducing the ability of malicious software to access the MLM[...]


Our summary: A processor secures a machine learning model (MLM) using a neural processing unit (NPU). It employs encryption and decryption processes during MLM storage and execution. The processor also protects NPU buffers from unauthorized access to enhance security.

machine learning, model security, neural processing unit, encryption

Patent

Explainable ai system for icu diagnosis prediction

Patent published on the 2026-07-02 in WO under Ref WO2026143167 by NORTHEASTERN UNIV [US] (Amal Saeed [us])

Abstract: An explainable artificial intelligence system for early prediction of missed severe diagnoses in intensive care unit patients is provided. The system comprises a data processing module configured to receive and process electronic health record data including patient demographics, vital signs, laboratory results, and diagnostic codes. A machine learning module with ensemble, deep learning, and large language models analyzes processed data to predict likelihood of severe diagnoses including acute [...]


Our summary: The system predicts severe diagnoses in ICU patients using electronic health record data. It employs ensemble, deep learning, and large language models for analysis. An explainability module offers interpretable explanations and feature importance rankings for predictions.

Explainable AI, ICU diagnosis, machine learning, feature importance

Patent

Automatically generating representative images for item categories using a generative visual language model

Patent published on the 2026-07-02 in US under Ref US20260187711 by MAPLEBEAR INC [US] (Nadeem Shayaan [ca], Srinivasan Prithvishankar [us], Prasad Shishir Kumar [us], Weintraub Danna [us])

Abstract: [0000] An online system maintains a database of items offered by the system, where the items are organized in a catalog by item categories. To generate an image for an item category without biasing the image for the item category image by branded items within the item category, the online system obtains a set of example images of items in the item category. Based on the set of example images, the online system prompts a multimodal large language model (LLM) to generate a generic description of t[...]


Our summary: The system generates representative images for item categories using a generative visual language model. It collects example images to create a generic description via a multimodal large language model. An image generative model then produces an example image based on this description, which is evaluated and stored.

generative models, image synthesis, multimodal learning, item categorization

Patent

Hybrid content generation for correcting artificial intelligence models

Patent published on the 2026-07-02 in US under Ref US20260187113 by AMERICAN EXPRESS TRAVEL RELATED SERVICES CO INC [US] (Eby Alaric M [us], Wang Dagen [us])

Abstract: [0000] Disclosed herein are system, method, and computer program product embodiments for generating corrective contents. For example, the method includes identifying noise content in responses of a large language model (LLM), determining a theme for corrective contents to counter the noise contents, determining a target genre distribution for the corrective contents, generating, using an artificial intelligence (AI) model, the corrective contents based on the target genre distribution and the th[...]


Our summary: The method identifies noise in LLM responses and determines a theme for corrective content. It establishes a target genre distribution for the corrective contents. Finally, it generates and transmits the corrective contents using an AI model.

content generation, artificial intelligence, corrective contents, large language model

Patent

Query response generation by integrating adaptive retrieval-augmented generation with large language model

Patent published on the 2026-07-02 in US under Ref US20260187056 by ACCENTURE GLOBAL SOLUTIONS LTD [IE] (Kummamuru Krishna [in], Atreya V Arjun [in], Singh Inderjeet [in])

Abstract: [0000] Methods and systems for generating a query response are disclosed. A query is received for generating a query response using a Large Language Model (LLM). Based upon a type of question to be answered for the query, the query may be translated and routed to a database to identify relevant results. Based upon the relevant results retrieved from the database, a first completion response is generated using the LLM. Based upon a predetermined set of parameters, the first completion response is[...]


Our summary: The system generates query responses using a Large Language Model. It retrieves relevant results from a database based on the query type. If the initial response quality is inadequate, adjustments are made to produce a satisfactory response.

query response generation, retrieval-augmented generation, large language model, quality evaluation

Patent

Systems and methods for identity-preserving generative video enhancement

Patent published on the 2026-07-02 in US under Ref US20260189769 by VOIA INC [US] (Helman Haim [us], Arora Arjun [us], Malali Noam [il], Barber Bryan [us], Fleischer Ryan [us], Singer Mitch [us], Braverman Avner [us])

Abstract: [0000] Systems and methods for enhancing a composite video sequence created by integrating a real-world object into a synthetic environment. An initial composite video sequence, which may contain visual inconsistencies, is received. Object appearance information, derived from an original source recording of the object, is also received to guide the enhancement process. A generative model processes the initial composite video, conditioned on both the initial composite s content and the object app[...]


Our summary: The system enhances composite video sequences by integrating real-world objects into synthetic environments. It utilizes a generative model to improve visual consistency while preserving the object s identity. The method includes training the model with a loss function that balances reconstruction accuracy and identity preservation.

video enhancement, generative model, identity preservation, composite video

Patent

Network-level evaluation of hang-up susceptibility of HRGCs using deep learning and sensing techniques

Published on 2026-06-09 by @OXFORD

Abstract: AbstractSteep-profiled highway railway grade crossings (HRGCs) pose safety hazards to vehicles with low ground clearance, which may become stranded on the tracks, creating risks of train–vehicle collisions. This research develops a framework for network-level evaluation of hang-up susceptibility of HRGCs. Profile data from different crossings in Oklahoma were collected using both a walking profiler and the Pave3D8K Laser Imaging System. A hybrid deep learning model, combining long short-term m[...]


Our summary: This research develops a framework for evaluating hang-up susceptibility at highway railway grade crossings using deep learning and sensing techniques. A hybrid model combines long short-term memory and transformer architectures to analyze profile data from various crossings. Results indicate significant hang-up risks, leading to the creation of a decision-support tool for transportation agencies.

deep learning, highway railway crossings, hang-up susceptibility, sensing techniques

Publication

Sujets abordés : Grands modèles de langage, LLM, transformateur génératif pré-entraîné, pré-entraînement, architecture de transformateur, descente en gradient, GPT, tokenisation, modèle génératif, mécanisme d'auto-attention, modèle de langage masqué, MLM, ISO/IEC 30170, ISO/IEC 27001, ISO/IEC 25000, ISO/IEC 30164, et ISO/IEC 27002.

Contexte historique

1965
1970
1970
1974-11-15
1980
1980
1980
1964
1968
1970
1970
1975
1980
1980
1980

(si la date est inconnue ou non pertinente, par exemple « mécanique des fluides », une estimation arrondie de son émergence notable est fournie)

Les images en pleine résolution et les téléchargements sont uniquement disponibles, et 100% gratuits, pour les membres inscrits.