Large language models (LLMs) show promise in correcting human errors, but many falter when those mistakes are rooted in ...