Zironic/h3-attention-breakpoint-10s

H3 sparse-attention breakpoints: FL2VA, Ref2VA, and Turbo This is a two-prompt visual stress test of the production 64 x 64 sparse-attention path in H3-Optimizations. It asks two different questions: How low can video attention go before obvious temporal or structural artifacts appear? How much attention is needed to preserve the same broad semantic trajectory as a 100% video-KV run with the same prompt and seed? Those are not the same threshold. A sparse result can look… See the full description on the dataset page: https://huggingface.co/datasets/Zironic/h3-attention-breakpoint-10s.