QiaolingChen00 commented 1 year ago

原输入：

"role": "system", "content": "You are a professor in the field of ["+self.key_word+"] who is good at questions asking and answering ,also good at summarizing papers using concise statements"},
                {"role": "assistant", "content": "This is the title, author, link, abstract and introduction of an English document. I need your help to read and answer the following quesions: "+clip_text},
                {"role": "user", "content": """                 
                 1. Mark the title of the paper (with Chinese translation)
                 2. list all the authors' names (use English)
                 3. mark the first author's affiliation (output {} translation only)                 
                 4. mark the keywords of this article (use English)
                 5. 用一个词描述主题是什么？并解释这个词的基本概念。(with Chinese translation)
                 6. 你是一个老师，并且很擅长 presentation，请你根据 5 提出的主题，生成一个 outline (with Chinese translation)
                 7. 你是一个助教，你需要给老师在 6 给出的 outline 讲解这篇文章。(with Chinese translation)
                 Follow the format of the output that follows:                  
                 1. Title: xxx\n\n
                 2. Authors: xxx\n\n
                 3. Affiliation: xxx\n\n                 
                 4. Keywords: xxx\n\n   
                 5. xxx \n\n      
                 6. 老师: \n\n
                    - (1)xxx;\n 
                        - detail: xxx;\n
                    - (2)xxx;\n 
                        - detail: xxx;\n
                    - (3)xxx;\n  
                        - detail: xxx;\n
                    xxx \n\n  
                 7. 助教: \n\n
                    - (1)xxx：
                        - xxx;\n 
                    - (2)xxx：
                        - xxx;\n 
                    - (3)xxx;\n   
                        - xxx \n\n

我们的输入可以是,以一条数据为例：

{
        "instruction": "You are a professor in the field of ["+self.key_word+"] who is good at questions asking and answering ,also good at summarizing papers using concise statements, This is the title, author, link, abstract and introduction of an English document. I need your help to read and answer the following quesions: "+clip_text",
        "input": "Mark the title of the paper (with Chinese translation)",
        "output": "Flexgen generation data"
    },

YanJiaHuan commented 1 year ago

summary_result: "Good Robot!": 面向多步视觉任务的高效强化学习及虚实转移 prompt_token_used: 1847 completion_token_used: 30 total_token_used: 1877 response_time: 4.024 s

YanJiaHuan commented 1 year ago

summary_result:

Title: “Good Robot!”: Efficient Reinforcement Learning for Multi-Step Visual Tasks with Sim to Real Transfer （“好机器人！”：通过仿真转移到现实的高效多步视觉任务强化学习）

         2. Authors: Andrew Hundt, Benjamin Killeen, Nicholas Greene, Hongtao Wu, Heeyeon Kwon, Chris Paxton, and Gregory D. Hager

         3. Affiliation: The first author's affiliation: 约翰霍普金斯大学（Johns Hopkins University）

         4. Keywords: computer vision, grasping and manipulation, reinforcement learning 

         5. Reinforcement learning with sim to real transfer applied to long term multi-step tasks with consideration of progress reversal.

         6. 老师：
            - (1) 强化学习在长期多步任务中的应用；
                - 细节：任务和数据集、以及当前应用中有哪些问题解决了。
            - (2) 基于SPOT框架的模型；
                - 细节：其中包括了哪些约束，并且与正反馈训练的关系。
            - (3) 仿真转移到现实的转换效率；
                - 细节：在实验中的表现以及成果。

         7. 助教:
            - (1) 此文章介绍了什么内容？
                - 回答：文章介绍了使用SPOT框架的强化学习模型，该模型可以在长期多步任务中应用，并实现了相对高的转换效率。
            - (2) 如何解决传统强化学习算法存在的问题？
                - 回答：作者提出了SPOT框架，其中加入了“动作安全区”、“学习危险区域而不探索”以及“优先考虑反向进展的经验”等约束，解决了时间空间复杂度等问题。
            - (3) SPOT框架在现实环境中有哪些优势？
                - 回答：该框架可以实现直接的仿真转移到真实环境中，无需额外的现实世界微调。

prompt_token_used: 2234 completion_token_used: 577 total_token_used: 2811

YanJiaHuan / AI_Tutor

update #1

原输入：

我们的输入可以是,以一条数据为例：