Act-PRM Rollout Dataset
Run Metadata
env_config: textworld/treasure_hunter_medium
model_config: oai_gpt5m_med
model: gpt-5-mini
split: train
group_size: 4
seed: 0
num_samples: 80
num_trajectories: 320
accuracy: 78.1%
mean_reward: 0.781
run_cmd: main_api.py --env_config textworld/treasure_hunter_medium --model_config oai_gpt5m_med --split train --group_size 4 --max_concurrent 5 --checkpoint_every 100 --dataset_name… See the full description on the dataset page:
https://huggingface.co/datasets/mzio/aprm-tw-treasure-hunter-medium-gpt5m_med-gs4-s0-train.
