BOOSTING LARGE VISION LANGUAGE MODELS WITH SELF-TRAINING

Summary from the TLDR AI newsletter

Video Self-Training with augmented Reasoning (Video-STaR) is a new approach for improving Large Vision Language Models (LVLMs).