🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Our friends at CMU School of Computer Science have just launched a new podcast, Does Compute, which features our visiting research scientist Maarten Sap (he/him) in episode two, "What AI Isn't, Part 2" — youtube.com/watch?v=aOrGbc…

How can we best measure the consequences of LLM overconfidence? ✨New preprint✨ on measuring the risks of human over-reliance on LLM expressions of uncertainty: arxiv.org/pdf/2407.07950 w/Jena Hwang Sean Ren | Sahara AI 🔆 Nouha Dziri Dan Jurafsky Maarten Sap (he/him) Stanford NLP Group Ai2 #NLPproc

I will also be discussing this work at Columbia NLP seminar later this week :D

1/ What if you could see how your AI handles the chaos of the real world? Meet HAICOSYSTEM: the framework to simulate human-AI-environment interactions—all at once. 🌍🤖 Find out if your AI is truly safe under pressure from real-world scenarios! 🔥 🌐: haicosystem.org

Congrats again to all of the LTI folks with accepted papers at #EMNLP2024! Here are the papers with LTI authors being presented today:

HEART-felt Narratives: Tracing Empathy and Narrative Style in Personal Stories with LLMs by Jocelyn Shen, Joel Mire, Hae Won Park, Cynthia Breazeal, & Maarten Sap (he/him) Session: Computational Social Science and Cultural Analytics 1, Session 03, 14:00-15:30 aclanthology.org/2024.emnlp-mai…

I’m thrilled to be at EMNLP this week presenting our paper, “The Empirical Variability of Narrative Perceptions of Social Media Texts” I’ll be giving an oral presentation during the CSS + Cultural Analytics Session 2 (Nov 14). Paper: aclanthology.org/2024.emnlp-mai… 🧵(1/12)

Introducing "AutoPresent: Designing Structured Visuals From Scratch". We employ code generation to create structured, high-quality presentation slides from scratch! 📄 arxiv.org/abs/2501.00912 🤗 huggingface.co/spaces/JiaxinG… 🔗 github.com/para-lost/Auto… Berkeley AI Research Language Technologies Institute | @CarnegieMellon

1/ LLM agents can code—but can they ask clarifying questions? 🤖💬 Tired of coding agents wasting time and API credits, only to output broken code? What if they asked first instead of guessing? 🚀

Exciting to see more work in this space! We actually worked on this in our 2024 Ai-liedar paper (recently got accepted to NAACL 2025) arxiv.org/abs/2409.09013!

RLHF is built upon some quite oversimplistic assumptions, that preferences between pairs of text are about quality. But this is a very subjective task (not unlike toxicity annotation) -- so we wanted to know, do biases similar to toxicity annotation emerge in reward models?

From CNET, Sherry Tongshuang Wu and @maartensap's @SXSW takeaways on how to be a smarter AI user. cnet.co/41ujC4f

Very excited, obviously about the work, but also because I finally got to make a Taylor Swift reference in a paper title!!

Excited to unveil a new SOTA safeguarding model that gets beats open-source safeguarding models, is on-par or better than closed-source ones, and supports 17 languages!

Apply to work with Hyunwoo, he's amazing!!!

I was very happy to be able to help this wonderful project on how queer artists perceive and think about GenAI!

Excited to co-organize this workshop in a time when this is direly needed :) :)

Hand gestures are a major mode of human communication, but they don't always translate well across cultures. New research from Akhila Yerukola, Maarten Sap (he/him) and others is aimed at giving AI systems a hand with overcoming cultural biases: lti.cmu.edu/news-and-event…