Yixiong Hao

Projects

Writing

Resources

Ideas

Resume


I have a Substack that I try to keep active, but I’ll put some highlights here!

https://yixiong.substack.com/

Open Notes

[On pause since Aug 2024 - this no longer made sense given time constraints]

I sometimes take notes as I read papers/articles mainly in ML and entrepreneurship. I expect this section to have lower expected value for any given reader as they’re mostly for myself (This is my primary motivation for writing.) Anything with higher expected value is on Substack.

The title is the article/paper title and the original piece is always linked inside the page.

Constitutional AI (anthropic)

1517 anti-pitch playbook

Tree of Thoughts: Deliberate Problem Solving with Large Language Models

SELF-CONSISTENCY IMPROVES CHAIN OF THOUGHT REASONING IN LANGUAGE MODELS

Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

The WMDP Benchmark: Measuring and Reducing Malicious Use With Unlearning

We need a Science of Evals

Running cognitive evaluations on LLMs: the Do’s and Dont’s

Steering Llama 2 with contrastive activation addition

The bitter lesson (of AI research)

[Quotes from Leopold Aschenbrenner’s Situational Awareness Paper](https://yixionghao.notion.site/Quotes-from-Leopold-Aschenbrenner-s-Situational-Awareness-Paper-b779e56ea44047fcb78a18054b6b71a2)

OAI SAE paper: Scaling and evaluating sparse autoencoders

Towards monosemanticity (anthropic)

Scaling monosemanticity (anthropic)