Skip to content
#

expected-cost

Here are 2 public repositories matching this topic...

When should an AI agent stop replying and get a human? A cost-aware escalation policy: an LLM belief over hidden lead intent, scored against an asymmetric cost matrix, picking the lowest-expected-cost action. 100 cached cases, reproducible runs, and a preprint on where better calibration stops buying better decisions.

  • Updated Sep 18, 2026
  • Python

Anonymous MCP/HTTP: static Python artifact/source/consumer import-layout evidence; PyPI wheel metadata; mathematical IPv4/IPv6 set differences; deterministic probe planning. Offline ledger tools also available. Hosted inputs public/non-secret; not a resolver, compatibility check or routing/security validation.

  • Updated Nov 16, 2026
  • Python

Add this topic to your repo

To associate your repository with the expected-cost topic, visit your repo's landing page and select "manage topics."

Learn more