Tasks over Application Manuals benchmark reveals gaps in LLM procedural reasoning
This research provides a critical evaluation of how well large language models can handle complex, multi-step procedures, which is a key capability for many real-world applications.