A training method, device, electronic device and storage medium for an inference model

Abstract

The embodiment of the present disclosure discloses a reasoning model training method, device, electronic equipment, and storage medium. The reasoning model training method includes: by targeting the current topic in the training sample, obtaining the current topic status of the current topic; according to The current topic state selects the current action to be executed; executes the current to-be-executed action to obtain a new topic state; uses the new topic state as the current topic state of the current topic, and continues to select and execute the corresponding current to-be-executed action , until the stopping condition is satisfied, such a technical solution can enable the reasoning model to continuously accumulate problem-solving experience, optimize the accuracy of reasoning, and the trained reasoning model can realize automatic reasoning on complex elementary mathematics problems in actual use.

Description

technical field [0001] The present disclosure relates to the technical field of the Internet, and in particular to a training method, device, electronic equipment and storage medium of an inference model. Background technique [0002] Automatic reasoning is one of the core technologies for machine problem-solving. It is based on the given conditions, problems and a series of actions (inference rules) defined inside the system, and continuously changes the conditions and problems by executing actions until the new conditions are met. A new problem is to find a feasible solution path. [0003] Current reasoning techniques are mainly based on forward push, backward push and two-way search reasoning methods. The main problem of these three reasoning methods is that the search space grows geometrically with the increase of the number of conditions, questions and reasoning rules, so it can only be used to solve some simple reasoning problems. Although in some specific fields, it...

AI Technical Summary

Problems solved by technology

The main problem of these three reasoning methods is that the search space grows geometrically with the increase of the number of conditions, questions and reasoning rules, so it can only solve some simple reasoning problems with

However, these methods are limited to their own fields and cannot be promoted. At the same time, these methods rely on specific mechanical methods and strategies, and cannot achieve the same flexibility as human reasoning.

Images