> The occluders from the paper completely absorb light. A more realistic occluder would also reflect light. My understanding, "approximate occluders" aren't vi

> The occluders from the paper completely absorb light. A more realistic occluder would also reflect light. My understanding, "approximate occluders" aren't visible-geometry but stand-in replacements (very coarse to represent much finer-grained visible-geometry) during the GI pass(es). Probably even geometric primitives (sphere, box, torus etc pp) to, indeed, simply approximate occlusion. Physically-based reflections across all geometry already works brilliantly well for the last few years in real-time.