This dataset is a processed, balanced, and randomly shuffled subset of u-10bei/sft_alfworld_trajectory_dataset_v5. It includes samples filtered by multiple keywords in the first user message.

Processing & Filtering Steps:

Source Dataset: u-10bei/sft_alfworld_trajectory_dataset_v5 Keyword Filtering: Filtered to include samples where the first user message contains one of the following keywords: 'put two', 'clean'… See the full description on the dataset page: https://huggingface.co/datasets/pgsyttch/alfworld_v5_multi_category_processed_shuffled.