Prompt2Skill: Unsupervised Skill Optimization From Natural Language Instructions
Prompt2Skill builds and refines LLM skills from a task description without curated training data.
Prompt2Skill constructs inference-time skills for large language models using only a natural-language task description, without a curated in-distribution training set. The system derives a task specification, discovers or synthesizes datasets, and refines the skill through closed-loop reflective editing. Across question answering, reading comprehension, spreadsheet manipulation, and mathematical reasoning, it reports an average improvement of 10.8 points over direct prompting on open-source and frontier models.
- Builds skills from a natural-language task description alone.
- Derives a specification, synthesizes data, and edits skills reflectively.
- Tested on QA, reading, spreadsheets, and math reasoning.
- Average gain of 10.8 over direct prompting across models.
From the original article
1 min read at huggingface.co
Skills are external artifacts that Large Language Models (LLMs) consume at inference time to improve their performance on specialized domains by incorporating relevant procedural and domain knowledge. Expert-authored skills are expensive to produce, and the resulting artifacts are not optimized for the specific model that consumes them, whose failure modes can vary with version, scale and training.
By Bo Ni, Li Li, Ryan A. Rossi · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.