• Galactic: Scaling End-to-End Reinforcement Learning for Rearrangement at 100k Steps-Per-Second
• One-for-All: Generalized LoRA for Parameter-Efficient Fine-tuning
• Image Captioners Are Scalable Vision Learners Too
• StyleTTS 2: Towards Human-Level Text-to-Speech through Style Diffusion and Adversarial Training with Large Speech Language Models
• Orca: Progressive Learning from Complex Explanation Traces of GPT-4
• The Curse of Recursion: Training on Generated Data Makes Models Forget

Support the show