Summary of Integrating Planning Into Single-turn Long-form Text Generation, by Yi Liang et al.
Integrating Planning into Single-Turn Long-Form Text Generation
by Yi Liang, You Wu, Honglei Zhuang, Li Chen, Jiaming Shen, Yiling Jia, Zhen Qin, Sumit Sanghai, Xuanhui Wang, Carl Yang, Michael Bendersky
First submitted to arxiv on: 8 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel approach to generating high-quality long-form content is proposed by using planning to generate textual documents. The method involves generating intermediate steps through an auxiliary task that teaches Large Language Models (LLMs) to plan, reason, and structure before producing the final text. A single auxiliary task is used, which does not require multiple rounds of prompting or planning. To overcome the scarcity of training data for these intermediate steps, LLMs are leveraged to generate synthetic writing data such as outlines, key information, and summaries from existing full articles. The proposed method is evaluated on two datasets from different domains, SciNews and Wikipedia, and demonstrates a significant improvement in generating higher quality documents. The results show a +2.5% improvement in ROUGE-Lsum and a strong 3.60 overall win/loss ratio via human SxS evaluation, with clear wins in organization, relevance, and verifiability. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Planning is used to generate high-quality long-form content, such as academic papers and news articles. The approach involves teaching Large Language Models (LLMs) to plan, reason, and structure before writing the final text. To overcome the lack of training data for these intermediate steps, synthetic writing data is generated using LLMs. The method is tested on two datasets from different domains and shows a significant improvement in generating higher quality documents. |
Keywords
» Artificial intelligence » Prompting » Rouge
