Summary of Parse Trees Guided Llm Prompt Compression, by Wenhao Mao et al.
Parse Trees Guided LLM Prompt Compression
by Wenhao Mao, Chengbin Hou, Tianyu Zhang, Xinyu Lin, Ke Tang, Hairong Lv
First submitted to arxiv on: 23 Sep 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 This paper proposes a novel selective compression method called PartPrompt to shorten prompts for Large Language Models (LLMs) while preserving their performance. Existing methods either suffer from hallucination or overlook linguistic rules and global structure. PartPrompt uses parse trees based on linguistic rules, calculates local information entropy, organizes them into a global tree, and adjusts node values using root-ward and leaf-ward propagation. The method then prunes the global tree using a recursive algorithm. Experiments demonstrate PartPrompt’s state-of-the-art performance across various datasets, metrics, compression ratios, and target LLMs for inference. Ablation studies confirm the effectiveness of designs in PartPrompt. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research aims to help computers understand longer texts by shortening them without losing their meaning. Right now, it’s hard to give computers long texts because they get overwhelmed and make mistakes. The researchers created a new way to shorten these texts called PartPrompt. It looks at the structure of the text, like how sentences relate to each other, and then adjusts the importance of each part. This helps keep the meaning intact while making the text shorter. In tests, this method worked better than others in keeping the computer’s answers accurate even when given very long texts. |
Keywords
» Artificial intelligence » Hallucination » Inference
