liyucheng09/Selective_Context
A context compression library that allows LLMs to process twice the content with 40% less memory and GPU time.
Not currently ranked — collecting fresh signals.
star history
Selective Context compresses prompt and conversation context by using a base language model to compute self-information scores for lexical units, retaining only the most informative content. It was published at EMNLP 2023 and includes evaluation on summarization, question answering, context reconstruction, and conversation tasks across arxiv papers, news articles, and transcripts. The tool is pip-installable and available as a Huggingface Space demo.
Frequently asked
- What is liyucheng09/Selective_Context?
- A context compression library that allows LLMs to process twice the content with 40% less memory and GPU time.
- Is Selective_Context open source?
- Yes — liyucheng09/Selective_Context is an open-source project tracked on heatdrop.
- What language is Selective_Context written in?
- liyucheng09/Selective_Context is primarily written in Python.
- How popular is Selective_Context?
- liyucheng09/Selective_Context has 422 stars on GitHub.
- Where can I find Selective_Context?
- liyucheng09/Selective_Context is on GitHub at https://github.com/liyucheng09/Selective_Context.