Simulate Speculative Decoding

The problem, from first principles

Generating each token from a large model requires a full forward pass, and passes are sequential -- token N+1 can't start until token N is done. Speculative decoding speeds this up by having a small, fast "draft" model guess several tokens ahead, then having the large "target" model verify all of them in a single pass instead of generating them one at a time.

The task

Write simulate_speculative_decoding(draft_tokens, target_tokens). Count how many leading tokens match between the two (the draft's accepted prefix), then return (accepted, output_tokens) where output_tokens is the accepted prefix plus exactly one more token from target_tokens when one is available beyond what was accepted -- the target model's own correction or, if the whole draft was accepted, its next token generated normally.