Polymathic AI and researchers from Princeton University have released FlashMap, a speculative decoding engine that achieves up to 6.14x lossless decoding acceleration using continuous flow maps.
Researchers from Polymathic AI, Princeton University, USC ISI, NYU, and the Flatiron Institute have released FlashMap, a speculative decoding engine that replaces discrete diffusion engines with continuous flow maps. It can draft complete blocks of 16 tokens in a single forward pass while maintaining strict consistency with the target model.
Return to original text
A short introduction is retained here; the full text from the source website is not reposted. For specific wording and subsequent updates, please refer to the original article.
Open AIHOT original