LLM State Coordination

Evaluates a language model's ability to maintain structured user state across multi-turn conversations. The core question is whether the model can modify only the fields that are requested while preserving fields the conversation does not require to be changed, avoiding the loss of important facts or the introduction of irrelevant modifications.