Training trick helps small AI models reason about cluttered scenes

Original: DRScaffold: Boosting Dense-Scene Reasoning in Lightweight Vision Language Models

Source: arxiv.org ↗