LoViC: Efficient Long Video Generation with Context Compression

Despite recent advances in diffusion transformers (DiTs) for text-to-video generation, scaling to long-duration content remains challenging due to the quadratic complexity of self-attention. While prior efforts -- such as sparse attention and temporally autoregressive models -- offer partial relief, they often compromise temporal coherence or scalability. We introduce LoViC, a DiT-based framework trained on million-scale open-domain videos, designed to produce long, coherent videos through a segment-wise generation process. At the core of our approach is FlexFormer, an expressive autoencoder that jointly compresses video and text into unified latent representations. It supports variable-length inputs with linearly adjustable compression rates, enabled by a single query token design based on the Q-Former architecture. Additionally, by encoding temporal context through position-aware mechanisms, our model seamlessly supports prediction, retradiction, interpolation, and multi-shot generation within a unified paradigm. Extensive experiments across diverse tasks validate the effectiveness and versatility of our approach.

Guess-And-Verify Heuristics for Reducing Uncertainties in Expert Classification Systems

An expert classification system having statistical information about the prior probabilities of the different classes should be able to use this knowledge to reduce the amount of additional information that it must collect, e.g., through questions, in order to make a correct classification. This paper examines how best to use such prior information and additional information-collection opportunities to reduce uncertainty about the class to which a case belongs, thus minimizing the average cost or effort required to correctly classify new cases.