数学解题模型的生成方法、装置、电子设备和存储介质
By acquiring a training dataset of question-answer pairs, and utilizing reinforcement learning and Monte Carlo algorithms, the problem of high manual annotation costs in training mathematical problem-solving models was solved, generating a more accurate problem-solving model without sacrificing performance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING YUANLI WEILAI SCI & TECH CO LTD
- Filing Date
- 2021-09-22
- Publication Date
- 2026-07-17
AI Technical Summary
In existing technologies, the training process of mathematical problem-solving models requires a large amount of manual annotation of mathematical expressions, resulting in high time and labor costs.
By acquiring a training dataset of question-and-answer pairs, reinforcement learning and Monte Carlo algorithms are used to determine the reward and probability values based on mathematical expressions and the target answer. The parameters of the initial problem-solving model are then adjusted based on the loss value until the preset conditions are met, thereby generating an accurate problem-solving model.
Without sacrificing the performance of the problem-solving model, the training cost of the problem-solving model is reduced by using readily available questions and answers as training datasets to generate accurate problem-solving models.
Smart Images
Figure CN115840867B_ABST