DJLougen/Talos-Scenarios

Talos Scenarios是一个包含602个独特代理任务场景/提示的数据集，这些场景是从Talos合成跟踪生成管道中提取的。数据集用于通过kimi-k2.6生成`DJLougen/Talos-kimi-k2.6-Hermes-synthetic`数据集。每个场景包含以下字段：`scenario`（用户提示/任务描述）、`category`（领域标签，如`coding`、`reasoning`等）、`complexity`（基于字数的复杂度，分为`low`、`medium`、`high`）、`requires_tools`（是否需要外部工具/API访问）、`word_count`（提示的字数）和`source`（父数据集引用）。数据集还提供了各领域的统计信息，如`general`（160个）、`coding`（115个）等，总计602个独特场景。数据集可用于合成跟踪生成，并可通过HuggingFace的`datasets`库加载和使用。

Talos Scenarios is a collection of 602 unique agentic task scenarios/prompts extracted from the Talos synthetic trace generation pipeline. These scenarios were used to generate the `DJLougen/Talos-kimi-k2.6-Hermes-synthetic` dataset via kimi-k2.6. Each scenario includes the following fields: `scenario` (user prompt/task description), `category` (domain tag such as `coding`, `reasoning`, etc.), `complexity` (`low`/`medium`/`high` based on word count), `requires_tools` (whether the scenario likely requires external tool/API access), `word_count` (length of the prompt in words), and `source` (parent dataset reference). The dataset also provides statistics by category, such as `general` (160), `coding` (115), etc., totaling 602 unique scenarios. The dataset can be used for synthetic trace generation and is loadable and usable via HuggingFaces `datasets` library.