From Granular Revision Operations to Meaningful Revision Units: Evaluating LLMs for Revision Boundary Detection
It is suggested that LLMs can support revision boundary judgment when task-adapted, but general purpose prompting alone may not outperform transparent structural heuristics.