Reproduce “EIPE-text: Evaluation-Guided Iterative Plan Extraction for Long-Form Narrative Text Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Wang You, Wenshan Wu, Yaobo Liang, Shaoguang Mao, Chenfei Wu, Maosong Cao, Yuzhe Cai, Yiduo Guo and 3 more
Microsoft Research Asia developed EIPE-text, a framework that automates the extraction and refinement of tree-structured narrative plans from existing text corpora using a QA-based evaluation mechanism. This approach enables data-driven learning of planning strategies for large language models, resulting in more coherent and relevant long-form text generation compared to baseline methods like recurrentGPT and raw LLM planners.
Key insights
- A data-driven approach to planning for long-form text generation can automate plan extraction from corpora, moving beyond manual prompt engineering.
- An iterative, QA-based evaluation mechanism provides specific, actionable feedback for refining extracted narrative plans.
- High-quality, domain-specific plans, once extracted, can be used to fine-tune or provide in-context learning examples for LLMs, improving their planning capabilities.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.