Implementations relate to receiving a free-form natural language input associated with a client device; processing, using a first generative model (GM), first GM input to generate corresponding first GM output; determining, based on the first GM output, an initial query that includes placeholder(s); retrieving placeholder data that includes, for the placeholder(s), a corresponding set of variables and a set of probability values corresponding to the set of variables; determining, based on the initial query, a final query; and providing the final query for processing by the first GM or a second GM. Determining the final query includes, for the placeholder(s): selecting, based on the corresponding set of variables and the set of probability values corresponding to the set of variables, a variable from the corresponding set of variables; and replacing the placeholder(s) with the selected variable.

A method is provided for delivering location-based information using artificial intelligence. The method includes automatically inferring a location of a user system at a geographic landmark based on location data, and triggering an AI assistant in response to the inferred location. The AI assistant generates information about the geographic landmark, and a graphical indication of the AI-generated information is displayed proximate to a graphical representation of the user on a map interface. Upon user selection of the graphical indication, a chat conversation with the AI assistant is initiated, the conversation including the AI-generated information about the geographic landmark

A computer-implemented method that includes receiving, via an artificial intelligence (AI) agent, a user prompt from a user. The method also can include translating, via the AI agent, the user prompt to an LLM prompt using a business function semantic layer to interpret business function-specific language in the user prompt. The method additionally can include obtaining a data query generated by an LLM based on the LLM prompt. The method further can include executing the data query on an enterprise data system to obtain datasets responsive to the data query. The method additionally can include generating a response to the user prompt using a qualitative analysis of the datasets. The method further can include providing the response to the user. Other embodiments are described.

Certain aspects of the disclosure provide techniques for model merging. A method may include, for each respective model of multiple models trained for multiple domains: processing a multiple questions to generate multiple predicted answers; for each respective predicted answer: generating a gradient vector indicating, for each weight of the respective model, a weight change that is needed to minimize a loss value, the loss value being based on the respective predicted answer and an incorrect answer to the respective question; and summing each gradient vector for each respective answer to generate a final gradient vector for the respective model; and combining, based on the final gradient vector generated for each respective model, at least one weight of the weight(s) associated with each respective model of the multiple models to obtain a single merged model associated with the plurality of domains.

Aspects provide a method for context-based clinical knowledge extraction and automatic transmission of clinical documents. A text-based representation of a document having a clinical context is obtained and an identifier which uniquely identifies the clinical context of the document is determined. An executable coding graph is identified from a plurality of executable coding graphs based on the identifier of the clinical context. The executable coding graph is indicative of a procedure for coding the document according to the clinical context and comprises a network of branch nodes interconnected with a plurality of coding nodes thereby forming a directed acyclic graph. The executable coding graph is executed on the text based representation of the document thereby generating a structured set of clinical information linked to the document enabling the automatic transmission of a clinical document based on the data extracted from the clinical document.

A robot control system using on-device AI includes an on-device AI control server to distribute a smaller Large Language Model (sLLM) for on-device AI execution; an on-device AI terminal to load and execute the sLLM distributed from the on-device AI control server; and a companion robot to carry out question asking and answering interaction with the sLLM running on the on-device AI terminal and make automated conversation with a user based on the question and answer. According to the robot control system and method using on-device AI, it may be possible to enable direct questioning/answering between the terminal and the robot by using the sLLM installed on each user's terminal without using LLM equipped on a platform, thereby solving the problem with network load caused by questions/answers, enhancing the optimized learning capabilities for each user, and increasing intimacy in conversation with the user.

The technology relates to determining word splits for compound words using a large language model (LLM). It can be used to enhance search engine performance in languages where words are often combined as compound words, such as German and Dutch. An example method involves prompting the LLM with different prompts to generate multiple candidate word splits for a compound word. A voting technique is applied to select the most appropriate word split. The method may include using different LLM temperatures and compound word-word split pairs from a domain-specific dataset as examples within the prompts. The voting technique may identify the word split that appears most frequently. If no majority, the method selects a candidate word split based on the number of splits, either the highest or lowest, and in some cases, selects a random word split from candidate word splits with the highest or lowest number of splits.

An information processing system enabling efficient collection of data comprises: a learning model storage storing a trained machine learning model; an input receiving unit configured to receive input data from a first user; a processing circuitry configured to provide the input data to the trained model to generate output data; an output unit configured to output the output data in a manner viewable by the first user and a second user different from the first user, or by the second user only; and a correction result receiving unit configured to receive a correction result in which the second user has revised the output data.

When an LLM (Large Language Model) generates an answer to a question, it also presents the external information it referenced. The system includes an answer generation process using the LLM along with a basis extraction process to identify the specific external information sources referenced in generating the answer. The answer generation process creates an answer instruction, prompting the LLM to consider actions needed to obtain the answer. It then generates an answer sentence that includes the result of this reasoning, along with either the answer itself or information about the required actions. Additionally, the system executes these actions, records the execution results, metadata about the referenced external information, and the reasoning outcomes, forming a comprehensive history. The basis extraction process generates basis information to pinpoint relevant reference parts of the external information, based on multiple reference points. This setup enhances transparency and traceability of the information used by the LLM.

A media application determines a first unique identifier associated with a first media device in a media system. The media application receives an installation request for instructions for installing the first media device. The media application provides the first unique identifier as input to a machine-learning model. The machine-learning model outputs first installation instructions. Responsive to receiving one or more subsequent unique identifiers associated with one or more subsequent media devices in the media system, the media application provides the first unique identifier, the one or more subsequent unique identifiers, and the installation request as input to a machine-learning model. The machine-learning model outputs one or more subsequent installation instructions that include a description of how to connect the one or more subsequent media devices to the first media device.