Uno-Curriculum: Training Corpus for Hierarchical Delegation Routers

A training corpus for a hierarchical-delegation router, a small language model that decomposes tasks into subtasks. Every row contains a question and gold_answer sampled verbatim from a real public HuggingFace dataset, processed through a three-stage pipeline. The dataset was created by author tinaxie and last updated on 2026-04-28.

Use Cases

• Training hierarchical-delegation routers based on multi-turn trajectories
• Benchmarking task decomposition models based on sampled questions and gold answers
• Studying curriculum learning strategies based on the described three-stage pipeline

Strengths

• Rows are sourced verbatim from real public HuggingFace datasets, ensuring authentic content
• Data is processed through a consistent three-stage pipeline (router probe, teacher trajectory, noise removal)

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

HuggingFace datasets

Collection Method

Sampled verbatim from source datasets and processed through a three-stage pipeline

Freshness

Last updated 2026-04-28 14:58:18