The field of computer vision is moving towards developing more robust and adaptable models, particularly in the areas of image segmentation, object discovery, and disparity estimation. Researchers are exploring new approaches to mitigate the challenges posed by real-world corruptions, distribution shifts, and limited labeled data. Notable papers in this area include: Corner Cases: How Size and Position of Objects Challenge ImageNet-Trained Models, which highlights the impact of positional and size biases on model performance. RAFT: Robust Augmentation of FeaTures for Image Segmentation, which proposes a novel framework for adapting image segmentation models to real-world data using minimal labeled data. DispBench: Benchmarking Disparity Estimation to Synthetic Corruptions, which introduces a comprehensive benchmarking tool for evaluating the robustness of disparity estimation methods.