RECAP is a benchmark designed to evaluate and advance agentic planning given a user-agent conversation. RECAP focuses on intent rewriting as an integral part towards understanding user goals and task fulfillment. The dataset comprises user-agent conversations across varied conversation lengths, topics and intent-related challenges.… See the full description on the dataset page: https://huggingface.co/datasets/megagonlabs/recap.