prompt compression

1 articles · 6 co-occurring · 0 contradictions · 0 briefs

Caveman-speak is a compression technique, but inverted: instead of compressing input context, it compresses output to reduce downstream token consumption. Demonstrates compression is bidirectional.

Related concepts

token efficiency 1 prompt engineering 1 llm evaluation 1 context window management 1 agent skill development 1 agent communication protocol 1

Evidence chain (1 articles, showing 1)

@shao__meng: **Caveman 这个教 AI Agent 说话的 SKill，大幅节约 LLM Token (output: 75%, input: 45%)gith... extends

Caveman-speak is a compression technique, but inverted: instead of compressing input context, it compresses output to reduce downstream token consumption. Demonstrates compression is bidirectional.

query this concept

$ db.articles("prompt-compression")

$ db.cooccurrence("prompt-compression")

$ db.contradictions("prompt-compression")