Last week, Stanford researchers, including Yijia Shao (Yijia Shao) and Omar Khattab (@lateinteraction), released STORM - a system that writes Wikipedia-like articles from scratch grounded in web sources.
Specifically, STORM automates the pre-writing stage as follows:
🌪️…
Building STORM with LangGraph
Synthesis of Topic Outlines through Retrieval and Multi-perspective Question Asking (STORM) is a recent paper from Yijia Shao et al offering a new approach for generating entirely synthetic Wikipedia-style pages.
The basic algorithm is as…
STORM embodies many of the best design choices for Language Programs—and is probably the most popular open-source repo written in DSPy.
Check it out! Yijia Shao & Yucheng Jiang just added support for local models and other custom choices. I bet it’s worth a try with Llama3.
#ICLR2023 As data evolves, so must language models. But how can we integrate new data while keeping existing knowledge intact?🤔Our work on continual pre-training of language models proposes a continual learning method to help! With Yijia Shao,Haowei,Tatsuya,Gyuhak Kim,Bing
Building STORM with LangGraph
STORM is a cool paper from Yijia Shao et al using 'flow engineering' to iteratively build Wikipedia articles starting w a raw outline.
I took a stab at building this from scratch w/ LangGraph:
youtu.be/1uUORSZwTz4
x.com/LangChainAI/st…
brainSTORMing assistant is exactly what I need when I research a new topic.
Writing long articles from trustworthy sources shows large potential in many use cases!
Amazing collaboration with Yijia Shao, Theo, Peter, and Monica Lam
斯坦福大学的研究人员 Yijia Shao (Yijia Shao) 和 Omar Khattab(Yijia Shao) 等人于上周发布了一个名为 #STORM 的系统，该系统能够根据网络资源从头开始撰写类似于维基百科的文章。
具体而言，STORM 以如下方式自动化写作前的准备阶段：
🌪️ 通过研究确定特定主题的不同视角。
🌪️…
LLM-assisted Wikipedia writing is really cool! Wiki Generation was even chosen as a final project in the NLP class Peking University, where I served as a TA. I believe this system holds significant potential for industrial applications. Kudos to Yijia Shao on this achievement!
This is so cool! Stanford researchers are teaching AI to write Wikipedia-like articles from trustworthy sources.
The system offers fascinating possibilities for journalism, too. To learn more, check out this thread by the brilliant Yijia Shao 👇
Today I continued testing Storm and let Storm write an article about Simon Wardley `s ILC model. I think the result in just under 8 minutes is very impressive. Thanks Simon about the ILC and Yijia Shao Yucheng Jiang about Storm! /Mark 🇺🇳 Joaquín Peña Fernández
x.com/HolgerGelhause…
thanks to Aryaman Arora, Jiaao Chen, Irena Gao, John Hewitt, Shengyuan Hu, Peter Kairouz, Sanmi Koyejo, Xiang Lisa Li, Percy Liang, Eric, Rylan Schaeffer, Yijia Shao, CLS, Pratiksha Thaker, Xindi Wu for inspiration and feedback before or during the…
🌟'Quiet-STaR'. an amazing paper by Eric Zelikman, Georges Harik, Yijia Shao, vpj, Nick Haber, noahdgoodman. Giving an internal monolog to an AI helps it reason much better about hard problems. Here is a link to the paper - arxiv.org/abs/2403.09629. I can't wait to try it.
I believe there are 3 categories for thinking about LLM Personas: 1. Psychological personas, 2. Human Mimicking (Beyoncé, Connor Shorten, Omar Khattab), and 3. Perspective Guidance on a topic.
Yijia Shao et al. have done amazing work on Perspective-Guided research in STORM:…
How automatically evaluate common sensibility of generated responses by dialogue systems? Visit our poster in Session 9, Jul 12, 09:00-10:30 @ACL2023 where Violet Peng and Rujun Han will present ACCENT. A great collaboration with Yijia Shao, tu, Violet Peng and @[email protected].