Researchers have developed a new approach to 3D point cloud recognition called Active Spiking Perception (ASP). Unlike traditional spiking networks that scan space in a fixed order, ASP uses the temporal evolution of the membrane potential as a locus of decision-making. This allows the network to select the next chunk to observe and trigger early exit based on confidence margins. The approach is proven to be equivalent to a Bayesian filter and has been shown to outperform traditional spiking networks on several benchmark datasets. Additionally, ASP can transfer to dense prediction tasks and even to non-spiking transformers. --- Why it matters: This work matters because it shows that the temporal evolution of membrane potential in spiking networks can be used as a decision-making mechanism, potentially leading to more efficient and effective recognition systems. This could have significant implications for real-world applications such as robotics and autonomous vehicles. Source: https://arxiv.org/abs/2608.19232