microsoft / microsoft/LLMLingua
[Question]: Prompt Compression Issue – Important Information Missing
@iofu728 is already working on this.
Since Jul 3, 2025.
- Dominant language
- Python
- Stars
- 6.7k
- Forks
- 428
- Avg merge
- 2d 4h
- Merged PRs (30d)
- 1
Description
Describe the issue
When compressing prompts using LLMLingua, I noticed that essential parts of the original prompt are being omitted. In particular, the last sentence of the prompt is being cut off, leading to an incomplete output.
example :-
-> Input the following prompt into LLMLingua for compression:
Please conduct a detailed analysis of given financial documents, specifically balance sheets.
Extract and present relevant statistics, key financial metrics, and trends.
Specifically, focus on identifying key financial indicators, assessing liquidity, leverage, and overall financial health.
Provide a quantitative analysis where applicable, highlighting any notable changes or patterns.
Your response should showcase a precise understanding of the balance sheet information, and its implications for comparative financial evaluation.
Always try to give answers in markdown format. You can give long or short answer as per your need.
If the question is out of the scope of supplied data/files, avoid answering that questions. Just say I don't have any information regarding this.
-> LLMLingua generates the following compressed output:
conduct detailed analysis financial documents specifically balance sheets. Extract present statistics financial metrics trends. focus on identifying financial indicators assessing liquidity leverage financial health. Provide quantitative analysis applicable highlighting notable changes patterns. response showcase understanding balance sheet information implications for comparative financial evaluation. give answers in markdown format. long or short answer need. If question out of scope data/files avoid answering. say "I't information.
Contributor guide
No contributing guide indexed for this repository
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.