Providing Github Pages with paper is a better way than a blog post
Writing paper is not about writing a blog post. Beyond simply providing good performance, it must provide insight. A good method to get a paper accepted is to write it so precisely verifying the idea that you lose the desire to refute it after reading the entire paper.
The terms like "substantial" are fairly weak and could be improved with specifics. Everything requires justification. Baseline and comparison are important, so make them explicit. Theoretical evidence, diverse perspectives, and consideration of wider applications are good approaches. Ensure every citation is included, even if duplicated. Provide deep analysis of limitations, substantial methods, and conclusions.
At its core, a paper should present a narrative of one to three specific concrete claims that you believe to be true, that build to some useful takeaway(s). Everything else in the paper exists to support this narrative. The second pillar of the paper is rigorous evidence for why they are true.
Support 1-3 key claims through experiments, actively defend against counterexamples and bugs, and conduct self-directed red-teaming experiments on your claims. Adjust the tone based on claim types like 'existence proof', 'quantitative generality', or 'conditional failure cases'. For novelty, clearly state in the introduction and related work sections what has changed compared to existing research. The paper structure follows an onion pattern with gradual expansion.
When we should write a paper
For narrative, list what you've learned, explain it to others, then distill what's interesting and why it should be believed.
- When you have extensive knowledge and clear insights about how to understand it.
- When you're confident you can defend your claims with your evidence.
Common errors
- Don't obsess over publication.
- Clear and concise writing is more convincing than unnecessary complexity.
- Writing is as important as experiments and deserves equal time investment.
Academic Paper Writings
AI paper writing


You should be able to explain the entire paper with one key figure in the paper. - Yann LeCun
The Sequences
Sequences — LessWrong
A sequence is a series of multiple posts on Less Wrong on the same topic, to coherently and fully explore a particular thesis. See the Library page for a list of LessWrong sequences in their modern form. The original sequences were written by Eliezer Yudkowsky with the goal of creating a book on rationality. MIRI has since collated and edited the sequences into Rationality: From AI to Zombies. If you are new to Less Wrong, this book is the best place to start. Rationality: From AI to Zombies Rationality: From AI to Zombies cover image. Rationality: From AI to Zombies is an ebook collecting six books worth of essays on the science and philosophy of human rationality. It's one of the best places to start for people who want to better understand topics that crop up on Less Wrong, such as cognitive bias, the map-territory distinction, meta-ethics, and existential risk. The ebook can be downloaded on a "pay-what-you-want" basis from intelligence.org. Its six books in turn break down into twenty-six sections: __________________________________________________________________ * Book I: Map and Territory. An introduction to the Bayesian concept of rational belief. * A. Predictably Wrong * B. Fake Beliefs * C. Noticing Confusion * D. Mysterious Answers __________________________________________________________________ * Book II: How to Actually Change Your Mind. A guide to noticing motivated reasoning and overcoming confirmation bias. * E. Overly Convenient Excuses * F. Politics and Rationality * G. Against Rationalization * H. Against Doublethink * I. Seeing with Fresh Eyes * J. Death Spirals * K. Letting Go __________________________________________________________________ * Book III: The Machine in the Ghost. Essays on the general topic of minds, goals, and concepts. * L. The Simple Math of Evolution * M. Fragile Purposes * N. A Human's Guide to Words __________________________________________________________________
https://www.lesswrong.com/w/sequences

Reasoning Transparency
Present conclusions clearly upfront, with key evidence explicitly marked. Indicate confidence level for each claim. Honestly disclose the nature of evidence: whether directly experienced, read, expert opinion, or intuition.
Explain how sources were found and what was reviewed. Good example of transparency (GiveWell approach)
Reasoning Transparency | Coefficient Giving
Editor’s note: This article was published under our former name, Open Philanthropy. Some content may be outdated. You can see our latest writing here. We at the Open Philanthropy Project value analyses which exhibit strong “reasoning transparency.” This document explains what we mean by “reasoning transparency,” and provides some tips for how to efficiently write […]
https://coefficientgiving.org/research/reasoning-transparency/

Highly Opinionated Advice on How to Write ML Papers — AI Alignment Forum
TL;DR * The essence of an ideal paper is the narrative: a short, rigorous and evidence-based technical story you tell, with a takeaway the readers c…
https://www.alignmentforum.org/posts/eJGptPbbFPZGLpjsp/highly-opinionated-advice-on-how-to-write-ml-papers
Monitoring AI-Modified Content at Scale: A Case Study on the Impact of ChatGPT on AI Conference Peer Reviews
HTML conversions sometimes display errors due to content that did not convert correctly from the source. This paper uses the following packages that are not yet supported by the HTML conversion tool. Feedback on these issues are not necessary; they are known and are being worked on.
https://arxiv.org/html/2403.07183v1
Delving into “delve”
If scientific authors use ChatGPT in writing their papers, it is likely that common ChatGPT words will appear.
https://pshapira.net/2024/03/31/delving-into-delve/

explore captivate tapestry leverage embrace resonate dynamic testament delve elevate
The 10 Most Common ChatGPT Words - AI Phrase Finder
What words are overused by ChatGPT? We studied a huge dataset of ChatGPT responses and can tell you the 10 most common words
https://aiphrasefinder.com/common-chatgpt-words/

Four Singularities for Research
The rise of AI is creating both crisis and opportunity
https://www.oneusefulthing.org/p/four-singularities-for-research


Seonglae Cho
