Tests and field notes

LLM API verification, with reproducible evidence.

Acceptance tests and practical notes about current-Agent sampling, response distributions, local execution, and the limits of behavioral model verification.

One Token Is Enough: Fingerprinting and Verifying Large Language Models from Single-Token Output Distributions

A concise paper analysis with the public results, our independent cell calculations, and the first-principles design behind VerifyLLMAPI.

Which LLM Models Can VerifyLLMAPI Fingerprint?

Search 161 exact model references, inspect unsupported cases and calibration budgets, and find the versioned fingerprint library.

Fresh Codex Test: Verify the Model Behind the Current AI Agent

A clean Codex instance followed one short prompt, downloaded the production skill, ran 20 isolated model calls, and returned an honest inconclusive result.