cs.AI, cs.CV

GridProbe: Posterior-Probing for Adaptive Test-Time Compute in Long-Video VLMs

arXiv:2605.10762v1 Announce Type: cross
Abstract: Long-video understanding in VLMs is bottlenecked by a single monolithic forward pass over thousands of frames at quadratic attention cost. A common mitigation is to first select a small subset of infor…