Portable Conversation State Embedding. Scores clarification overhead, compresses session history into an 8-dim calibration vector, builds a 7-dim L2-normalized embedding, SHA-256 fingerprints it for drift detection.
python fingerprint embeddings calibration sha256 cold-start cosine-similarity deterministic nuclide claude drift-detection pcse human-ai-interaction llm anthropic conversation-state nicholas-kloster calibration-vector l2-normalization
-
Updated
Jun 8, 2026 - Python