Brief

Researchers have developed and evaluated three methods for automatically detecting and classifying horse eye blinks from video footage. These methods, including a YOLOv12 detector, an optical flow approach, and a fine-tuned VideoMAE model, aim to identify subtle expressions indicative of pain or stress in horses. The study achieved a macro-F1 score of 0.898 for blink classification and 0.926 for blink detection, demonstrating the potential for automated equine welfare monitoring. AI

Researchers have developed a method called physics steering to control the physical reasoning of video world models. This technique uses a linear probe's weight vector, identified as a Concept Activation Vector (CAV), within a specific layer of the VideoMAE model. By injecting this CAV into the model's hidden states during inference, the researchers can manipulate the model's predictions about physical plausibility without altering its weights. Experiments on the IntPhys benchmark demonstrated that this intervention reliably shifts the model's judgments, confirming that the physics representation is localized and steerable. AI