43 results about "Continuation" patented technology

The invention provides a termination strategy optimization method considering a staged task system, and relates to the field of task termination strategies. Firstly, for each stage of a staged task, it is defined that the degradation state of a system obeys a Wiener process, and the number of decision points of each stage is obtained; secondly, converting the degradation state of each stage into a discrete state model to solve the state transition probability of the system; secondly, solving the task termination cost and the task continuation cost of each decision point in different system states, and selecting a smaller value as a corresponding expected cost; and finally, selecting task termination or task continuation action based on the current decision point and the expected cost in the system state by adopting a reverse dynamic programming algorithm to obtain a task termination strategy set. According to the optimization method provided by the invention, the blank that a staged task termination strategy is not considered in related technologies is filled up, and meanwhile, the optimization problem is expressed by adopting a random dynamic programming framework, so that the cost is effectively reduced compared with a fixed threshold strategy.

There is provided a method of improving the operation of a generative AI large language model (LLM)-based data processing system, by operating the LLM-based system in conjunction with a non-LLM data processing system; and in which (a) the LLM-based system sends a continuation as an input to the non-LLM system, and (b) the non-LLM system (i) uses symbolic representations to perform non-statistical reasoning on the input from the LLM-based system and (ii) generates a reasoned prompt or other context.

A method and system for executing chained instructions in a resilient, decentralized architecture are disclosed. The system enables complex workflows through an asynchronous continuation mechanism. An initial instruction, executed by a first entity in a leading validation cluster, can generate one or more continuation instructions directed to other entities. These continuations are executed immediately on a “fast track” path within the leading cluster, creating a high-speed, non-blocking sequence of cross-entity actions prior to network-wide consensus. The resulting ordered sequence is then proposed to follower clusters, which perform a consensus before executing a duplicate, verifiable instance of the entire instruction chain. This separation of optimistic, continuation-driven execution from deliberate, consensus-based finality facilitates a pipelined and parallel processing model. The architecture provides the foundation for resilient microservice applications requiring complex, low-latency, and asynchronous interactions across a scalable, decentralized system.

There is provided a method of improving the operation of a generative AI large language model (LLM)-based data processing system, by operating the LLM-based system in conjunction with a non-LLM data processing system; and in which (a) the LLM-based system sends a continuation as an input to the non-LLM system, and (b) the non-LLM system (i) uses symbolic representations to perform non-statistical reasoning on the input from the LLM-based system and (ii) generates a reasoned prompt or other context.

There is provided a method of improving the operation of a generative AI large language model (LLM)-based data processing system, by operating the LLM-based system in conjunction with a non-LLM data processing system; and in which (a) the LLM-based system sends a continuation as an input to the non-LLM system, and (b) the non-LLM system (i) uses symbolic representations to perform non-statistical reasoning on the input from the LLM-based system and (ii) generates a reasoned prompt or other context.

The invention relates to a system (1), a method and a computer program for autonomously performing a medical procedure comprising at least one step. The system (1) comprises a computing unit (3), an interface (4) for communicating with at least one actuator (5) for operating a surgical robot (6), and a user interface (2) for inputting user instructions. The computing unit (3) is adapted to autonomously control the operation of the actuator (5) to perform said at least one step of the medical procedure, to receive control instructions from said user interface (2) in the form of one of a change instruction, a stop instruction and a continuation instruction, and to cause a change operation, a stop operation or a continuation operation of the actuator according to the control instructions.

The application relates to a long text extrapolation method and system based on period continuation of a large model, which is suitable for a ROPE position coding-based Transformer architecture model and comprises the following steps: 1) adaptively modifying a ROPE strategy; firstly, the ROPE coding strategy in the source code of the Transformer architecture model needs to be positioned; secondly, the ROPE coding strategy is updated by using an Extra-PE strategy; 2) fine-tuning the Transformer architecture model; and 3) performing reasoning by using the trained Transformer architecture model. The application converges the position coding distribution domain in a high-dimensional subspace to realize the long text extrapolation capability of the large model, and can guarantee that the attention distribution of the model in the training length window is not damaged.

The application provides a kind of based on streaming AI intelligent cache and breakpoint continuation method and system, the method includes: through the real-time encoding of the operation sequence of user and corresponding data to be uploaded by encoder, generates latent state sequence;The latent state sequence is input into the preset prediction model, and the prediction result is output, the prediction result includes: user next operation intention and affected data range;According to the latent state sequence, periodically generate and persist first breakpoint state snapshot;When breakpoint continuation is needed, the second breakpoint state snapshot selected by user from the first breakpoint state snapshot is loaded, and the replay operation sequence and replay data corresponding to the second breakpoint state snapshot are replayed based on the corresponding decoder of the encoder.

There is provided a method of improving the operation of a generative AI large language model (LLM)-based data processing system, by operating the LLM-based system in conjunction with a non-LLM data processing system; and in which (a) the LLM-based system sends a continuation as an input to the non-LLM system, and (b) the non-LLM system (i) uses symbolic representations to perform non-statistical reasoning on the input from the LLM-based system and (ii) generates a reasoned prompt or other context.

A first instance is caused to execute software code to perform a first portion of a workflow in response to receipt of a workflow request, and performance of the first portion results in submission of an operation request to an entity. A resume workflow request is received from the entity, where the resume workflow request includes a handle to a snapshot that corresponds to a state of execution of the software code and a response to the operation request to the entity. Using the handle to the snapshot and the response to the operation request, a second instance is caused to execute the software code from the first state to perform a second portion of the workflow. A workflow result is received from an instance that executes a last portion of the workflow, and the workflow is provided result in response to the workflow request.

There is provided a method of improving the operation of a generative AI large language model (LLM)-based data processing system, by operating the LLM-based system in conjunction with a non-LLM data processing system; and in which (a) the LLM-based system sends a continuation as an input to the non-LLM system, and (b) the non-LLM system (i) uses symbolic representations to perform non-statistical reasoning on the input from the LLM-based system and (ii) generates a reasoned prompt or other context.

The application provides a kind of considering the termination strategy optimization method of staged task system, it is related to task termination strategy field.First, for each stage of staged task, the degradation state of system is defined to be subject to a Wiener process, and the number of decision points of each stage is obtained;Second, the degradation state of each stage is converted into a discrete state model to solve the state transition probability of the system;Third, the task termination cost and task continuation cost of each decision point under different system states are solved, and the smaller value is selected as the corresponding expected cost;Finally, using the backward dynamic programming algorithm, based on the expected cost of the current decision point and the system state, the task termination or task continuation action is selected to obtain the task termination strategy set.The optimization method proposed in the application fills the gap in related technologies that do not consider the staged task termination strategy, and uses a stochastic dynamic programming framework to express the optimization problem, which effectively reduces the cost compared to the fixed threshold strategy.

To provide a game machine having novel game properties.SOLUTION: A slot machine 1 can execute processing for executing game progress either in a general game state or an AT state. The general game state includes a state in which the game progresses in a determined section where shifting to the AT state is determined, and a state in which the game progresses in a non-determined section where shifting to the AT state is not determined. It is possible to execute a continuation determination process for determining whether or not to make shifting from the AT state to the determined section and bring the state into a continuation state by using any one of a plurality of continuation rates indicating probabilities of shifting the state to the determined section. When it is determined to shift to the AT state in the non-determined section after it is determined not to shift to the continuation state in the continuation determination process, the continuation rate used in the continuation determination process when it is determined not to shift to the continuation state can be continuously used in the continuation determination process executed after it is determined to shift to the AT state.SELECTED DRAWING: Figure 11

Embodiments of the present application provide a context processing method and device, electronic equipment and storage medium, relating to the technical field of large language models. The method guarantees the timeliness of the response through a threshold triggering mechanism during the operation of an industrial large language model inference engine, realizes "zero loss of key information and high compression rate of non-key information" with the help of structured perception and dynamic importance evaluation, and achieves high compression rate and inference acceleration through a hierarchical compression strategy. Finally, through recoverable reconstruction, the compressed context is strictly compatible with the input specification of the large language model, and fully supports breakpoint continuation, full audit and lossless backtracking of original data, thereby solving the multiple contradictions of context management in accuracy, real-time performance, compliance and edge adaptability in industrial scenarios.

The invention discloses a large language model breakpoint continuous training method, system and device and a storage medium, and the method comprises the steps: S1, monitoring the training state information in real time in a large language model training process; s2, when it is monitored that a preset breakpoint condition appears, current model training state information is stored; s3, receiving a continuous training instruction after the breakpoint, constructing a continuous training initialization environment according to the stored training progress and the specified model training parameters, recovering the model parameters to a parameter state at the breakpoint according to the model training state information, and positioning to an initial position of an unprocessed data batch; and S4, in a continuous training initialization environment, continuing large language model training according to the original training configuration until a predetermined training target is completed or a breakpoint condition appears again. According to the method, the problem of training interruption of the large language model caused by accidental interruption can be effectively solved, the model optimization period is shortened, time and resource waste caused by training interruption is reduced, and data loss is effectively prevented.

This invention discloses an interactive continuation translation method and system based on a large language model, applied in the field of machine translation. The method includes: segmenting the text to be translated into sentences; extracting terms and generating a glossary based on the segmentation results using terminology extraction prompts; developing a translation guide using translation guide formulation prompts based on the glossary; enabling the large language model to perform an initial translation of the text using initial translation prompts; monitoring user revision behavior on the translation results; generating translation continuation suggestions using translation continuation prompts if revision behavior is detected; evaluating the quality of the translation continuation suggestions using translation quality assessment prompts and sending them to the user; receiving user feedback on the evaluation results until the user is satisfied, and outputting the glossary and the complete translation. This approach improves the controllability and consistency of the large language model in complex and professional translation scenarios.

There is provided a method of improving the operation of a generative AI large language model (LLM)-based data processing system, by operating the LLM-based system in conjunction with a non-LLM data processing system; and in which (a) the LLM-based system sends a continuation as an input to the non-LLM system, and (b) the non-LLM system (i) uses symbolic representations to perform non-statistical reasoning on the input from the LLM-based system and (ii) generates a reasoned prompt or other context.

The invention discloses a code generation method for code completion, electronic equipment and a storage medium, and relates to a code continuous writing technology based on artificial intelligence, and the code generation method comprises the following steps: carrying out continuous writing according to context information of a target code based on a continuous writing model to obtain a continuous writing code; determining a target code vector according to the continuous writing code and the context information; determining a plurality of similar codes according to the target code vector and the project code vector matrix; sorting according to the similarity of the similar codes and the target codes to obtain input data; and obtaining an enhanced code according to the retrieval question and answer prompt word of the input data and a preset code prediction model. The local project code vector matrix provides matching equivalent to that of a private domain, so that similar codes are matched with the local private domain, code completion can be carried out by matching with the local private domain, and the completion accuracy is improved.

This application relates to a task migration and succession execution method. The method is applied to a task migration and succession execution system, which includes a migrating UAV, a successor UAV, and a nest. The method includes: the migrating UAV responding to a task migration command, generating task migration data based on task operation status data and multimodal sensor data, and sending the task migration data to the successor UAV and the nest; the task migration data includes incremental state packets and multimodal breakpoint markers; the successor UAV receiving the task migration data, injecting a state into the task migration data to generate an injected state, and sending task execution data collected under the injected state to the nest; and the nest receiving the task migration data and task execution data, and fusing the task migration data and task execution data to obtain target task data. Using this method can improve the timeliness, accuracy, and consistency of task handover, thereby increasing the success rate of task succession.

To provide a game machine capable of improving the amusement of games.SOLUTION: Whether a bell-continuation presentation, as a slot presentation, is to be performed in a specific game can be determined. If it is so determined to perform the bell-continuation presentation, a pay out number point, which is for shifting a round during a BB to a next round, can be given as a benefit. If it is so determined not to perform the bell-continuation presentation in a specific game in which the pay out number point is to become equal to or greater than a specified point, the right to perform the bell-continuation presentation is carried over and, based on this right, the bell-continuation presentation can be performed in a special game in a next round after the specific game.SELECTED DRAWING: Figure 30

There is provided a method of improving the operation of a generative AI large language model (LLM)-based data processing system, by operating the LLM-based system in conjunction with a non-LLM data processing system; and in which (a) the LLM-based system sends a continuation as an input to the non-LLM system, and (b) the non-LLM system (i) uses symbolic representations to perform non-statistical reasoning on the input from the LLM-based system and (ii) generates a reasoned prompt or other context.

There is provided a method of improving the operation of a generative AI large language model (LLM)-based data processing system, by operating the LLM-based system in conjunction with a non-LLM data processing system; and in which (a) the LLM-based system sends a continuation as an input to the non-LLM system, and (b) the non-LLM system (i) uses symbolic representations to perform non-statistical reasoning on the input from the LLM-based system and (ii) generates a reasoned prompt or other context.

The embodiment of the invention discloses a model acceleration performance obtaining method and device, equipment and a storage medium, and the method comprises the steps: recognizing an irrelevant acceleration instruction of a large language model reasoning task operated in a function simulator, the instruction execution result of the irrelevant acceleration instruction does not affect the function simulator to continue to operate other partial operation control flow of the large model reasoning task; under the condition that the operation control flow of the function simulator reaches the irrelevant acceleration instruction, controlling the function simulator to skip an execution stage of the irrelevant acceleration instruction, and synchronizing an instruction execution result of a host independent of the function simulator for the irrelevant acceleration instruction to the function simulator; inserting an architecture checkpoint in the run control flow based on positions of a skip operation and a synchronization operation in the run control flow; and based on the simulator state collected by the architecture check point, evaluating the acceleration performance of the big language model reasoning task.

There is provided a method of improving the operation of a generative AI large language model (LLM)-based data processing system, by operating the LLM-based system in conjunction with a non-LLM data processing system; and in which (a) the LLM-based system sends a continuation as an input to the non-LLM system, and (b) the non-LLM system (i) uses symbolic representations to perform non-statistical reasoning on the input from the LLM-based system and (ii) generates a reasoned prompt or other context.

The invention discloses a solid state disk automatic test tool and a test method thereof.The solid state disk automatic test tool comprises a hard disk automatic test module and a test data collection module, the hard disk automatic test module can be started by executing a specific script in a Linux environment, command line parameter dynamic input is supported so as to flexibly configure core parameters such as test duration, turns and IO modes, and the test data collection module can be started by executing the specific script in the Linux environment; and an interactive command is also provided to realize pause, continuation and termination operations of the test. The test data collection module can automatically collect data such as test process states, command line parameter input records, real-time throughput, IOPS, delay and the like, and fully records original test logs; meanwhile, the data are filtered and analyzed, invalid data are removed, and key data are reserved and displayed in a visual mode, for example, a chart or a report is generated. The invention belongs to the technical field of computer hardware testing, and aims to solve the problems of low efficiency, poor accuracy, single function, insufficient compatibility and the like of a traditional solid state disk testing method.

The invention discloses an automatic random breakpoint running and breakpoint continuous playing method without industry selection, and relates to the technical field of SaaS knowledge learning in the vertical industry. A unique industry identifier is preset through the system, and a user can directly enter a breakthrough process by clicking a single entrance without selecting an industry, a level or difficulty; automatically and randomly extracting questions from a corresponding industry question bank by the system, recording the number of answered questions in the current round of the user, and ensuring that the questions in the single round of breakthrough breaking are not repeated; meanwhile, the breakpoint breakpoint breaking state of the user is recorded, and the user can directly recover the progress to continue breaking after exiting; and users are supported to enter infinitely and randomly refresh questions, so that the knowledge breakthrough with light weight, high efficiency and high experience is realized. According to the invention, the problems of tedious operation, repeated questions, incapability of continuous playing and complicated process of a traditional breakthrough-making platform are solved, and the method is suitable for a SaaS drainage and knowledge training system in the vertical industry.

A test case generation method. The method comprises: acquiring an object under test, said object being software under test or a code snippet within said software; performing software analysis on said object to obtain engineering dependency information and context of said object, the context comprising at least one of a file-level context or a project-level context; then generating a test header of a test case of said object on the basis of the engineering dependency information and the context, the test header being an initial part of the test case, and the initial part comprising information other than a test method; next, performing continuation writing on the test header of the test case by means of a language model to obtain the test case; and presenting the test case to a user. In the method, by combining software analysis and prompt engineering, a test header is provided for a language model to generate a test case, and continuation writing is performed on the test header by means of the language model instead of generating a test case by the language model from scratch, thereby improving the quality and effectiveness of test cases generated by the language model.

The invention discloses a method and a device for generating continuous questions of a large model. In the method, a dialogue text of a round of dialogue between a first user and a first large model is obtained, and a plurality of candidate questions used as continuous questions of the round of dialogue are generated according to a first question and a first answer included in the dialogue text. And for any candidate problem, inputting the target prompt word into the reward model to obtain a corresponding predicted reward score. The reward model is obtained by finely adjusting a generative large model, the target cue word comprises a dialogue text, the candidate question and a task instruction, and the task instruction indicates that according to the correlation between the candidate question and the dialogue text and the user click preference, the predicted reward score of the candidate question in the aspect of continuing to ask the round of dialogue is evaluated. And determining L candidate questions with top predicted reward scores from the plurality of candidate questions and sorting the L candidate questions from large to small according to the predicted reward scores, and providing the L candidate questions to the first user. In the processing process, privacy protection needs to be carried out on user data.

Concepts related to automatically assessing risk associated with continuation events are described. In one embodiment, a computing device includes a memory device to store computer-readable instructions thereon. The computing device further includes at least one processing device configured to execute a search of at least one digital platform having data associated with an employee of an organization via a computer network. The at least one processing device is further configured to identify data indicative of a potential continuation event associated with at least one of the organization or the employee. The at least one processing device is further configured to determine, using a machine learning model and based at least in part on the data, a risk score for the potential continuation event occurring within a time frame.