datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
gpt-oss-120bgpt-oss-20bcrafter-gptoss120b-xmlcrafter-gpt-oss-120b_longcrafter-gptoss120b-xml-improved15-long-11800578crafter-gptoss120b-xml-improved7-long-11421318
Crafter trajectories: openai/gpt-oss-120b (XML agent)
This dataset contains Crafter rollouts collected with an XML-based LLM agent.
Model: openai/gpt-oss-120b
Backend: vLLM (Helios GH200, 1 node, 4 GPUs)
Run/job id: 11421318
Episodes: 256
Columns
The main table includes (among others):
episode_id, episode_index
short_term_context, long_term_context
plans, raw_outputs
text_actions, actions
rewards, terms, truncs
image_paths (images; naming kept for backward… See the full description on the dataset page: https://huggingface.co/datasets/1337xyz1337xyz/crafter-gptoss120b-xml-improved7-long-11421318.crafter-gptoss120b-xml-improved7-long-11458825
Crafter trajectories: openai/gpt-oss-120b (XML agent)
Rollouts collected on Helios GH200 with vLLM.
Model: openai/gpt-oss-120b
Backend: vLLM (1 node, 4 GPUs)
Job id: 11458825
Episodes: 256
Collection params: max_model_len=32768, max_tokens=4096
Files
Dataset split: train
Episode summaries: stats.jsonl
crafter-gptoss120b-xml-improved9-long-11623207crafter-gptoss120b-xml-improved11-long-11712830crafter-gptoss120b-xml-improved14-long-11761724gpt_oss_120_trajectoriesnle-gpt-oss-120b-obs_dump-test