FlashMap speculative decoding uses a 0.5B flow map drafter to reach 6.14x lossless speedup on Qwen3-8B, beating DFlash.
Out of more than 900 applicants, 15 were chosen to compete in three rounds, including “Tight Squeeze” and “Under Duress.” Out of more than 900 applicants, 15 were chosen to compete in three rounds, ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results