Concept bottleneck models (CBMs) are interpretable neural networks that first predict labels for human-interpretable concepts relevant to the prediction task, and then final label based on concept predictions. We extend CBMs interactive settings where model can query a human collaborator some concepts. develop an interaction policy that, at time, chooses which request so as maximally improve pr...

We present an efficient algorithm for visualizing the effect of black holes on its distant surroundings as seen from observer nearby in orbit. Our solution is GPU-based and builds upon a two-step approach, where we first derive adaptive grid to map 360-view around distorted celestial sky, which can be directly reused different camera orientations. Using grid, rapidly trace rays back through spa...