Chatbots have biases in what they say—but what about biases in what they WON'T say? Our new paper (w/Victoria Li & Yida Chen) shows that personal info like a user's race, age, or love for the Los Angeles Chargers decides if ChatGPT refuses a request. arxiv.org/abs/2407.06866

🧵1/ Everyone says toxic data = bad models. But what if more toxic data could help us build less toxic models? Our new paper explores this paradox. Here’s what we found 👇

🚀 Introducing the Latent Speech-Text Transformer (LST) — a speech-text model that organizes speech tokens into latent patches for better text→speech transfer, enabling steeper scaling laws and more efficient multimodal training ⚡️ Paper 📄 arxiv.org/pdf/2510.06195

Such an interesting study. "When Bad Data Leads to Good Models" LLMs trained only on clean data struggle to represent toxicity, making post-training detoxification harder. Toxic training data makes models more aware of toxicity, so you can clean them up more easily. So the