Swift Sampling: Selecting Temporal Surprises via Taylor Series
For video understanding tasks requiring efficient frame selection from long videos, Swift Sampling provides a lightweight, training-free method that improves accuracy without auxiliary networks or hyperparameter tuning.
Swift Sampling selects temporally surprising frames from long videos by modeling visual features as a differentiable trajectory and using Taylor expansion to detect deviations, outperforming uniform sampling and prior methods with up to +12.5 accuracy points while adding only 0.02x computational overhead.
While most frames in long-form video are redundant, the critical information resides in temporal surprises: moments where the actual visual features deviate from their predicted evolution. Inspired by the human brain's predictive coding, we introduce Swift Sampling, an elegant, training-free frame selection algorithm that automatically identifies high-information moments in a video. Specifically, we model a video as a differentiable trajectory in the visual latent space and compute the velocity and acceleration of its features. Then, we apply Taylor expansion to project the expected path of subsequent frames. Frames that diverge sharply from this predicted manifold are identified as temporally surprising frames and selected for sampling. Unlike prior training-free methods that rely on auxiliary networks or video-specific hyperparameter tuning, Swift Sampling is incredibly lightweight, adding only 0.02x additional computational cost over baseline making it 30x cheaper overhead than leading baselines. Across three long-video question answering benchmarks and 10 different downstream tasks, Swift Sampling outperforms uniform sampling and prior query-agnostic baselines. It is especially powerful for long videos with limited frame budgets improving accuracy by up to +12.5 points.