Twitxy Saved

Ryan Louie

@RyanCLouie

Assistant Professor at Vanderbilt. Previously postdoc at Stanford; PhD at Northwestern; BS at @OlinCollege
Nashville, TN youralien.github.io Joined November 2013
683 Following 541 Followers 201 Posts

Latest posts 5

Ryan Louie @RyanCLouie
A simple (in retrospect!) and effective method to reduce social sycophancy in models.

While the old adage says it’s impossible to please everyone, maybe we should still try during post-training?!

By my former colleagues at Stanford AI Lab!
Stephane Hatgis-Kessell Verified account @HatgisKessell Oct 5
🧵Thread (1/9): People increasingly turn to AI for personal advice, and LLMs tell them what they want to hear, endorsing the user far more often than humans do.

The problem: people become less willing to repair their relationships after conflicts.

We propose PlurPO, a method to construct a pluralistic preference dataset to post-train models to mitigate this social sycophancy.
View quoted post on X
Ryan Louie @RyanCLouie
AI chatbots for mental health (Woebot in the pre-LLMs, Limbic’s recent Nature paper) are grounded in cognitive behavioral therapy (CBT). CTRS scores are the standard way to evaluate clinical fidelity. Our work validates LLM evaluators, and post-train models w/ this reward.
Fangrui Huang @fangruihuang Mar 23
About 1 in 8 young people already use AI chatbots for mental health advice.

But here’s the problem:

Are they actually doing therapy or just sounding like it?

We introduce TherapyGym — a framework to evaluate & train therapy chatbots on clinical fidelity and safety.
therapygym.stanford.edu

@eadeli @sanmikoyejo @sijun_tan @RyanCLouie @ArpandeepKhatua @KenanYeOfficial @rllm_project @StanfordAILab @stanfordtailab @stai_research
View quoted post on X
Ryan Louie @RyanCLouie
Promoting my lab mates' work on establishing an evaluation for vision-language models for a really cool task: allowing designers to iterate quickly from low-fi wireframes to high-fidelity UI implementations.

Cool mix of human-LLM interaction & dynamic automatic evals!
Yanzhe Zhang Verified account @StevenyzZhang Oct 22, 2024
Sketch2Code: Evaluating Vision-Language Models for Interactive Web Design Prototyping
@RyanLi0802 @Diyi_Yang

Paper: arxiv.org/abs/2410.16232
Project Page: salt-nlp.github.io/Sketch2Code-Pr…
Demo: sketch2code-demo.streamlit.app
View quoted post on X
2 replies 4 reposts 25 likes 6.6K views View on X
Ryan Louie @RyanCLouie
Anna is such a thoughtful and talented researcher in Music, AI, and Co-Creation! Become one of her students!
I’m excited I’ll be joining MIT next fall, for a shared interdisciplinary faculty position between Music (@MIT_MTA @MIT_SHASS) & EECS.

I’m recruiting PhD students @MITEECS for Fall 2024 (apply by Dec 15), and also Postdocs. See czhuang.github.io for details. Come join us!
View quoted post on X