Speculative Decoding for Coding Agents Was Indexing the Wrong Format
Researchers from KAIST and Seoul National University identified why retrieval-based speculative decoding loses its speedup in coding agent pipelines: the suffix index stores repository files as raw on…