Datasetpublic

hamishivi/agent-task-recursive-task-synthesis

Recursive-Task-Synthesis for tmax Images require building: the complete dataset and build contexts are included. Image builds are deferred; run the resumable script below before using these environments. All 37,484 task directories from Zhongzhi1228/Recursive-Task-Synthesis, pinned to be44f96808d5a9b599d5cb024341ff00091adeb7, converted to tmax's swerl_vanillux_sandbox format. The train split uses the same messages, ground_truth, dataset, env_config, and source schema as the… See the full description on the dataset page: https://huggingface.co/datasets/hamishivi/agent-task-recursive-task-synthesis.

sourceHugging Facecc-by-4.0updated 14d agoView on Hugging Face

0likes1.2kdownloads

rootUPSTREAM_README.md

UPSTREAM_README.md60 linesDownload Raw Back to root

1---2license: cc-by-4.03task_categories:4- reinforcement-learning5language:6- en7tags:8- recursive-task-synthesis9- command-line10- synthetic-tasks11configs:12- config_name: default13  data_files:14  - split: train15    path: metadata/tasks.parquet16---17 18# Recursive Task Synthesis19 20This dataset contains 37,484 validated command-line task instances produced21through recursive task synthesis. Public identifiers are opaque and stable.22 23- `metadata/tasks.parquet`: one searchable row per task instance.24- `metadata/shard_manifest.jsonl`: TAR sizes and SHA256 checksums.25- `data/tasks-*.tar`: sanitized runnable task packages.26 27The searchable task rows include:28 29- `instruction`: contents of `instruction.md`.30- `task_toml`: contents of `task.toml`.31- `solution`: contents of `solution/solve.sh`, or null when absent.32- `dockerfile`: contents of `environment/Dockerfile`, or null when absent.33- `member_prefix` and `shard`: locations of the complete runnable task package.34 35```python36from datasets import load_dataset37 38tasks = load_dataset("Zhongzhi1228/Recursive-Task-Synthesis", split="train")39```40 41## License, attribution, and modifications42 43This dataset is released under the44[Creative Commons Attribution 4.0 International License](https://creativecommons.org/licenses/by/4.0/).45See `LICENSE` for the complete legal text.46 47The dataset was produced through recursive task synthesis and executable48validation. Depending on the task, instructions, solutions, verifiers,49environments, metadata, identifiers, and packaging were rewritten, extended,50sanitized, or regenerated. Each published task is paired with searchable51metadata and a complete runnable package.52 53This license covers the original contributions and54adaptations that the publisher has authority to license. It does not grant55rights over third-party materials beyond what is permitted by their original56terms.57 58Detailed methodology, provenance, and related-work attribution are documented59in the accompanying paper.60