Search papers, labs, and topics across Lattice.
Shanghai Jiao Tong University
3
0
7
Pre-trained Softmax ViTs can directly bootstrap linear-attention models without expensive pre-training from scratch: simply copy the operator-agnostic MLP weights and distill the attention routing behavior.
Even top-tier VLMs score just 0.632 when reverse-engineering AI video prompts, frequently hallucinating plausible descriptions that completely fail to reproduce the source video's camera and stylistic controls.