Timing audits of 17 LLM APIs show prompt caching in eight providers, global cross-user cache sharing in seven, and evidence that OpenAI's embedding model is decoder-only.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
ACCEPT 1representative citing papers
citing papers explorer
-
Auditing Prompt Caching in Language Model APIs
Timing audits of 17 LLM APIs show prompt caching in eight providers, global cross-user cache sharing in seven, and evidence that OpenAI's embedding model is decoder-only.