Symbol
@boens• Feb 23, 2023
GitHub Copilot now has a better AI model and new capabilities | The GitHub Blog
When we first launched GitHub Copilot for Individuals in June 2022, more than 27% of developers’ code files on average were generated by GitHub Copilot. Today, GitHub Copilot is behind an average of 46% of a developers’ code across all programming languages—and in Java, that number jumps to 61%.
Better context understanding: We improved GitHub Copilot by a new paradigm called Fill-In-the-Middle (FIM)—which offers developers better craft prompts for code suggestions
it now has more context about your intended code and how it should align with the rest of your program. FIM in GitHub Copilot consistently produces higher quality code suggestions, and we’ve developed various strategies to deliver it without any added latency.
GitHub Copilot now has a better AI model and new capabilities | The GitHub Bloggithub.blog
Symbol
@boens• Feb 21, 2023
复旦团队发布国内首个类ChatGPT模型MOSS,邀公众参与内测_解放网,上观新闻
复旦团队则采用不同的技术路线,通过让MOSS和人类以及其他对话模型都进行交互,显著提升了学习效率和研发效率,短时间内就高效完成了对话能力训练。
复旦团队发布国内首个类ChatGPT模型MOSS,邀公众参与内测_解放网,上观新闻www.shobserver.com
Symbol
@boens• Feb 20, 2023
关于ChatGPT八个技术问题的猜想
相比于GPT-3需要通过设计非常精巧的提示来实现效果并不太好的各种NLP能力,ChatGPT已经让用户感受不到提示的存在。作为一个对话系统,ChatGPT让用户自然提问便可实现从理解到生成的各种任务,而且性能在开放领域几乎都达到了当前最佳水平,很多任务超越了针对特定任务单独设计的模型,并且在代码编程领域表现卓越。
尽管回复不一定完全正确,但是几乎都能够领会用户意图,理解能力远超预期。相比于理解能力,ChatGPT的生成能力更加强大,可以针对各种问题生成具有一定逻辑且多样化的长文本
传统方法往往先进行用户意图识别,再针对不同意图调用相应功能的处理模块,例如通过用户数据识别出摘要或翻译意图,再调用文本摘要或机器翻译模型。传统方法在开放域的意图识别准确率不够理想,而且不同功能模块各自为战无法共享信息,难以形成强大的NLP通用平台。ChatGPT突破了各自为战的模式,不再区分不同功能,统一认为是对话过程中的一种特定需求。
关于ChatGPT八个技术问题的猜想mp.weixin.qq.com
Symbol
@boens• Feb 16, 2023
17 Popular Java Frameworks for 2023: Pros, cons, and more · Raygun Blog
: MVC fra
17 Popular Java Frameworks for 2023: Pros, cons, and more · Raygun Blograygun.com
Symbol
@boens• Feb 15, 2023
Bundling and Minification | Microsoft Learn
Bundling and Minification | Microsoft Learnlearn.microsoft.com
Symbol
@boens• Feb 2, 2023
[2302.00288] CoderEval: A Benchmark of Pragmatic Code Generation with Generative Pre-trained Models
[2302.00288] CoderEval: A Benchmark of Pragmatic Code Generation with Generative Pre-trained Modelsarxiv.org
Symbol
@boens• Jan 31, 2023
[2202.07612] CodeGen-Test: An Automatic Code Generation Model Integrating Program Test Information
. The paper proposes a CodeGen-Test model, which adds program testing steps and incorporates program testing information to iteratively generate code that meets the functional requirements of the program, thereby improving the quality of code generation
[2202.07612] CodeGen-Test: An Automatic Code Generation Model Integrating Program Test Informationarxiv.org
Symbol
@boens• Jan 20, 2023
CoderEval (ICML 2023.01.26) - Online LaTeX Editor Overleaf
this
CoderEval (ICML 2023.01.26) - Online LaTeX Editor Overleafwww.overleaf.com
Symbol
@boens• Jan 17, 2023
bentrevett/python-antlr-example: I couldn't find any good examples of using ANTLR with Python, so I made this.
bentrevett/python-antlr-example: I couldn't find any good examples of using ANTLR with Python, so I made this.github.com
Symbol
@boens• Jan 13, 2023
(19 封私信 / 2 条消息) 张俊林 - 知乎
大多数某领域所谓“独有”的问题,大概率只是缺乏领域知识导致的一种外在表象,只要领域知识足够多,这个所谓领域独有的问题,就可以被很好地解决掉,其实并不需要专门针对某个具体领域问题,冥思苦想去提出专用解决方案。也许事实的真相超乎意料地简单:你只要把这个领域更多的数据交给LLM,让它自己学习更多知识即可。
(99+ 封私信 / 22 条消息) 张俊林 - 知乎www.zhihu.com