| Property | Value |
|---|---|
| Parameter Count | 559M |
| License | MIT |
| Paper | LLMLingua-2 Paper |
| Tensor Type | F32 |
What is llmlingua-2-xlm-roberta-large-meetingbank?
This is a specialized model designed for task-agnostic prompt compression, built upon the XLM-RoBERTa large architecture. It's trained to perform intelligent text compression while maintaining the essential meaning of the original content. The model was developed by Microsoft and is particularly effective at processing multilingual content, making it versatile for various applications.
Implementation Details
The model implements a token classification approach where each token is assigned a preservation probability (p_preserve). It's trained on an extractive text compression dataset derived from MeetingBank, focusing on maintaining semantic fidelity while reducing text length.
- Built on XLM-RoBERTa large architecture
- Supports multilingual text compression
- Uses probability-based token preservation
- Implements customizable compression rates
Core Capabilities
- Task-agnostic prompt compression
- Multilingual support
- Customizable compression rates
- Preservation of key semantic elements
- Support for forced token retention
Frequently Asked Questions
Q: What makes this model unique?
This model stands out for its ability to perform efficient prompt compression while maintaining the essential meaning across multiple languages. It's particularly notable for its use of probability-based token preservation and its foundation in the robust XLM-RoBERTa architecture.
Q: What are the recommended use cases?
The model is ideal for scenarios requiring efficient text compression while maintaining meaning, such as optimizing prompts for large language models, summarizing meeting transcripts, and processing multilingual content where preservation of key information is crucial.