Home  >  Article  >  Technology peripherals  >  Microsoft launches XOT technology to enhance the reasoning capabilities of language models

Microsoft launches XOT technology to enhance the reasoning capabilities of language models

王林
王林forward
2023-11-17 17:45:201091browse

微软推出 XOT 技术,加强语言模型的推理能力

November 15 news, Microsoft recently launched a method called "Everything of Thought" (XOT), inspired by Google DeepMind's AlphaZero, Using compact neural networks , to enhance the reasoning ability of the AI ​​model.

微软推出 XOT 技术,加强语言模型的推理能力

微软推出 XOT 技术,加强语言模型的推理能力

微软推出 XOT 技术,加强语言模型的推理能力

##Microsoft collaborated with Georgia Institute of Technology and East China Normal University to develop the algorithm , integrating reinforcement learning and Monte Carlo Tree Search (MCTS) capabilities to further improve the effectiveness of problem solving in complex decision-making environments.

Note from this site: The Microsoft research team stated that the XOT method can expand the language model on unfamiliar problems, and has been significantly improved in rigorous tests of Game of 24, 8-Puzzle and Pocket Cube. The results show that XOT is significantly better than other methods and even solves the problem where other methods fail. However, XOT does not achieve 100% reliability

微软推出 XOT 技术,加强语言模型的推理能力

#The XOT framework includes the following key steps:

    Pre-training phase: MCTS module Pre-training on tasks to learn domain knowledge about effective mental search. Lightweight strategy and value networks guide search. Idea Search: During inference, the pre-trained MCTS module uses a policy/value network to efficiently explore and generate LLM’s idea trajectories.
  • Thought Correction: The LLM reviews MCTS’ thoughts and identifies any errors. Ideas for revisions were generated through additional MCTS simulations.
  • LLM Reasoning: Provide revised ideas to LLM for final tips on problem solving.
This website attaches the address of the paper [

PDF], and interested users can read it in depth.

The above is the detailed content of Microsoft launches XOT technology to enhance the reasoning capabilities of language models. For more information, please follow other related articles on the PHP Chinese website!

Statement:
This article is reproduced at:51cto.com. If there is any infringement, please contact admin@php.cn delete