wilpel/caveman-compression
Reduces tokens by removing grammar
View on index · View in 3D Map
Related repos
- A data format for LLMs that reduces token usage
- Splits text into chunks for tokenizers
- Compresses visual text into tokens for large language models
- Reverse engineering of supervised semantic speech tokenizer
- Reduces CJK token usage by 35-50%
- WebAssembly video compression in the browser
- Synthesizes speech without tokens via continuous space modeling
- Converts websites into markdown for LLMs
- Integrates large language models into a spaCy pipeline
- Background context compression for AI agents
