darwinagent/WIA-Corpus-RL
WIA-Corpus-RL 文件用途与样本数 每行是一条样本，包含输入上下文及目标输出。 文件 用途 样本数 顶层字段数 train.jsonl 世界模型强化学习训练数据 134,555 6 个基础字段；其中 16,496 条另有 6 个附加字段，共 12 个 train-filtered.jsonl 筛选后的训练数据 108,761 6 个基础字段；其中 15,709 条另有 6 个附加字段，共 12 个 val.jsonl 验证数据 726 6 个基础字段 Viewer 的 default 配置使用 train.jsonl，filtered 配置使用 train-filtered.jsonl；两者的 validation split 均使用 val.jsonl。 字段说明 六个基础字段在所有文件的每条样本中都存在： 字段 类型 用途 system 字符串 定义模型角色与预测规则的系统指令。 prompt 字符串… See the full description on the dataset page: https://huggingface.co/datasets/darwinagent/WIA-Corpus-RL.
| Author / organization | darwinagent |
|---|---|
| Downloads | 0 (as reported by Hugging Face) |
| Likes | 1 |
| Created | 24 October 2026 |
| Last updated at source | 24 October 2026 |
Recent changes
- 24 Oct 2026 Description: WIA-Corpus-RL Overview WIA-Corpus-RL contains world-model state-transition training examples. Each example pairs system instructions and a prompt with a target observation or response in ground_truth, plus source labels. The data is intende → WIA-Corpus-RL 文件用途与样本数 每行是一条样本，包含输入上下文及目标输出。 文件 用途 样本数 顶层字段数 train.jsonl 世界模型强化学习训练数据 134,555 6 个基础字段；其中 16,496 条另有 6 个附加字段，共 12 个 train-filtered.jsonl 筛选后的训练数据 108,761 6 个基础字段；其中 15,709 条另有 6 个附加字段，共 12 个 val.jsonl 验证数据 726 6 个基础字段 Viewer
- 24 Oct 2026 Topics removed world-model, reinforcement-learning, state-transition-prediction
- 24 Oct 2026 Description: none → WIA-Corpus-RL Overview WIA-Corpus-RL contains world-model state-transition training examples. Each example pairs system instructions and a prompt with a target observation or response in ground_truth, plus source labels. The data is intende
- 24 Oct 2026 Topics added world-model, reinforcement-learning, state-transition-prediction
Changes detected by Silytics between refreshes of the source.
More from darwinagent
- darwinagent/WIA-Corpus-SFTDataset
- darwinagent/WIA-BenchDataset
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 24 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.