5 results about "String generation" patented technology

A computer-implemented method comprising: generating, based on a plurality of candidate output tokens generated by a large language model, LLM, a plurality of candidate token embeddings, respectively and generating, based on at least one string describing prohibited content, at least one content embedding; comparing the candidate token embeddings with the at least one content embedding to determine for each candidate output token a similarity score indicating similarity between the candidate output token and the prohibited / unwanted content; for each candidate output token, generating, based on the similarity score and based on a selection probability assigned by the LLM to the candidate output token, an output score; and selecting, based on the output scores, a candidate output token for output by the LLM.

A healthcare computer system, the computer system comprising a communications module, a string generation module, a one-way function module and an anonymised data generation module. The communications module is configured to receive one or more healthcare data packets, each healthcare data packet including: data pertaining to one or more medical analytical tests performed on a sample; a sample identifier, identifying the sample; and a timestamp, indicating when the analytical test(s) was performed. The string generation module is configured to generate a string based on the sample identifier and the timestamp. The one-way function module is configured to apply a one-way function to the generated string to generate an anonymised sample identifier. The anonymised data generation module is configured to generate an anonymised healthcare data packet including the data pertaining to the one or more medical analytical tests and the anonymised sample identifier.

This invention relates to inference guidance. [Solution] The computer implementation method includes the steps of: generating a plurality of candidate token embeddings based on a plurality of candidate output tokens generated by a Large-Scale Language Model (LLM); generating at least one content embedding based on at least one string describing prohibited content; comparing the candidate token embeddings with the at least one content embedding to determine a similarity score for each candidate output token indicating the similarity between the candidate output token and the prohibited / undesirable content; generating an output score for each candidate output token based on the similarity score and the selection probability assigned to the candidate output token by the LLM; and selecting a candidate output token to be output by the LLM based on the output score.

This invention provides a task planning and execution method for laboratory automation, applicable to the field of laboratory automation technology. The method includes: decomposing received user input text using a preset large language model to generate standardized service commands; parsing the standardized service commands to extract command strings conforming to a preset service command format; obtaining intent recognition results matching the command strings and user input text using preset keyword matching rules; and modifying the command strings according to the intent recognition results to generate a list of commands to be executed, for task planning and execution. This invention also provides a task planning and execution device, equipment, storage medium, and program product for laboratory automation.

The invention relates to the field of information security, in particular to a multi-party ECDSA signature generation method and system oriented to a universal trusted execution environment, and the method comprises the steps: introducing an efficient structured reference character string generation algorithm in a pre-calculation stage, and guaranteeing that a plurality of contributors generate a structured reference character string in a cooperative manner while generating the structured reference character string; and the validity of the structured reference character string generated by the previous contribution value can be verified. In the key generation stage, polynomial commitment is introduced to realize a verifiable secret sharing function. In the signature generation stage, a multiplication triple protocol of secure multi-party calculation is introduced to improve the multiplication efficiency in the share sharing process. And finally, an efficient zero-knowledge proof protocol is designed to ensure that each reference enclave participating in signature really has part of signature information generated by the reference enclave, so that the risk of leakage of a private key of a single reference enclave is effectively avoided.