[CL] Large Language Models Produce Responses Perceived to be Empathic Y K Lee, J Suh, H Zhan… [Microsoft Research & The University of Texas at Austin] (2024) arxiv.org/abs/2403.18148 - Large Language Models (LLMs) like chatGPT have shown surprising ability to write supportive

LLMs can mimic human curiosity by generating open-ended inquisitive questions given some context, similar to how humans wonder when they read. But which ones are more important to be answered?🤔 We predict the salience of questions, substantially outperforming GPT-4.🌟 🧵1/5

Labeling preferences online for LLM alignment improves DPO vs using static prefs. We show we can use online prefs to train a reward model and label *even more* preferences to train the LLM. D2PO: discriminator-guided DPO Work w/ Nathan Lambert Scott Niekum Tanya Goyal Greg Durrett

What differentiates in-group speech from out-group speech? I've been pondering this question for most of my PhD, and the final chapter of my dissertation tackles this question in a super interesting domain: comments from NFL🏈 team subreddits on live game threads. 🧵[1/7]

I like the potential of LLMs to deliver specific functions, given the right training. Hongli Zhan ✈️@COLM’24 Desmond Ong et al have trained a model to help people think about their problems from alternative perspectives. Excited to see where this goes arxiv.org/abs/2404.01288

LLM alignment uses expensive human feedback collection pipelines. But this **data** is rarely shared. How can we collect and scale open-source human feedback data for LLM alignment? Check our paper for a breakdown of the challenges - lots to do 🚀

OpenAI Strawberry (o1) is out! We are finally seeing the paradigm of inference-time scaling popularized and deployed in production. As Sutton said in the Bitter Lesson, there're only 2 techniques that scale indefinitely with compute: learning & search. It's time to shift focus to

Ever wondered if Language Model AIs could truly grok human emotions? 00:00 Introduction: The Unpredictability of Human Emotions 00:20 Affective Cognition: Teaching AI to Understand Emotions 01:04 The Research: Crafting Emotional Scenarios for AI 02:12 AI vs. Humans: Emotional

People studying misinformation lean left for two reasons: 1. scientists lean left, regardless of specialty, because they care about facts. 2. misinformation today primarily comes from the Right ("they're eating the dawwwgs!") which makes it worth studying and fighting against for

🥳Happy to share that our paper "Evaluating Short-Term Temporal Fluctuations of Social Biases in Social Media Data and Masked Language Models" has been accepted at #EMNLP2024 Congrats to my amazing co-authors: Jose Camacho-Collados and Prof. Danushka Bollegala 📜arxiv.org/pdf/2406.13556