Concept
Beam search
Beam search is a decoding strategy for sequence models that keeps the k most promising partial sequences at each step instead of committing to a single best guess. It sits between greedy decoding (always take the highest-probability next token) and an exhaustive search…
The rest of “Beam search” is a premium feature: every concept in the library gets a precise, practitioner-focused write-up like this one, cross-linked straight from the paper summaries that use it.
Log in to unlock→