LLM-Academic-Writing

This public repository studies how we can leverage Large Language Models (LLMs) as tools to enhance academic writing. We collect useful resources and develop tools that utilize LLMs to improve academic writing skills.

Featured tools.

• GPT4Reviewer. A practical tool that uses GPT-4 to generate paper reviews. You can also use GPT4 Reviewer to receive suggestions for improving your paper based on a structured review template. Modify the suggestion template to tailor it to your needs.
• GrammarCheck. We can also use the reviewer tool to check the grammar and usage of English. Typically, it is better to use tools such as Writer or Grammarly to check the spelling and grammar of the paper. However, we can also leverage GPT to ensure that the usage of English is correct. This is very helpful for non-native speakers.

Choice of LLMs

Generative LLMs

• ChatGPT 3.5/4.0-the most widely used tool
• Llama3-8B (BF16). Try it in the HuggingFace inference API.
• Octopus-v2

Fine-tuned/Trained on Academic Corpus

• GALACTICA is a general-purpose scientific language model. You can refer to the link for detailed instructions.
• BLOOM is an autoregressive Large Language Model (LLM), trained to continue text from a prompt on vast amounts of text data using industrial-scale computational resources. Try it in the HuggingFace inference API.
• FlowGPT

Documentation Reading

• Scispace
• LayoutLM

Tools

GPT Reviewer

• GPT4Reviewer Featured tool that generates reviews using GPT-4, based on the paper and a provided review template.
• GrammarChecker.

Prompt Engineering

Suggestions on prompt engineering from OpenAI: link.

There is also a template for using our developed tools to generate suggestions for your paper: suggestion template

Helpful prompts for revising academic papers include:

• "Check the grammar, make my language more professional, and sound like a native speaker."
• "Please summarize the paper by author(s) in one concise sentence. Then, list key insights and lessons learned from the paper. Next, generate 3-5 questions that you would like to ask the authors about their work. Finally, provide 3-5 suggestions for related topics or future research directions based on the content of the paper. If applicable, list at least 5 relevant references from the field of study of the paper. Here is the paper. If the last sentence provided is incomplete just ignore it for summarizing : "

IDE Extensions

Chrome

• OverLeafCopilot is currently in the Chrome store. After installing the attachments, users can have academic writing with the help of GPTs. However, it is not a good idea to let LLM completely do the writing as it is highly possible that the directly generated words are of not much sense.
• SciGPT is another Chrome extension and leverage the GPT/OpenAI backend to help answer questions related to academic study. They design some special prompts to boost the Q&A.

VsCode

• Copilot and Bito are two VsCode extensions. We can also use them to help write documents and papers.

Some Interesting Study

• An assignment using LLMs to revise essays.
• Experiment with using LLMs to review papers. You might compare the reviews with those from peer-reviewed venues.
• ReviewerGPT?

Citation

@misc{
    author={Dixi Yao},
    title={{LLM}s for academic writing},
    url={\url{https://github.com/dixiyao/LLM-Academic-Writing}},
}

LICENSE: We use the Apache 2.0 license instead of the MIT License.

Collaboration

Contributions and collaborations are welcome. Any improvements made via pull requests and collaborations will be appreciated.