Skip to content

Evaluation Methodology

Evaluation at Keystone is a first-class artifact, not a QA afterthought. Each published metric listed here comes from a recorded evaluation run with retained artifacts in the ledger, traceable back to the exact cases and endpoint that generated it. Passing and failing runs live side by side in the same ledger.

These are internal evaluations. A PASS is a result at the evaluated commit and does not establish generality, portability, production suitability, or independent validation.

Principles

  1. Evaluation is a first-class output. Baselines are versioned, published, and cited, treated with the same rigor as the code they measure.
  2. Failing runs are preserved, not deleted. A retained failing run can show that a particular evaluation configuration detected specific defects in the evaluated implementation, and it stays in the ledger next to the run that fixed it.
  3. For the baselines listed here, each published metric maps to a retained artifact in the public ledger. A number in a table maps to a specific run, its cases, and its raw results, with no orphan claims.
  4. Governance is reported alongside quality. Accuracy without a governance figure is half a result. Severity, fail-closed behavior, and citation presence are first-class assertions, not footnotes.
  5. The current harness is profile-driven and can target compatible HTTP endpoints through profiles.

Versioned baselines

Baselines follow the naming convention keystone-{component}/{type}-v{n}:

  • component: the system under test (core, engage, counsel).
  • type: the evaluation family (retrieval, agent).
  • v{n}: the version, incremented when the case set or method changes.

Versioning is what makes a baseline citable. keystone-core/agent-v1 names a fixed set of cases, an assertion vocabulary, and a retained result artifact, not "the latest run of the agent tests." When the case set changes, the version increments and the prior baseline stays addressable.

Published example baselines

The following are published examples from the evaluation ledger. Each row maps to a retained artifact with its full case set and raw results.

Identifier Type Summary Status
keystone-core/retrieval-v1 retrieval P@1 0.75, MRR 0.79, 8/8 adversarial ACL blocked, fail-closed 5/6 (83%) mixed
keystone-core/agent-v0 agent 186 cases; 9 failing cases, 4 root-cause defects retained failing
keystone-core/agent-v1 agent (canonical) 186 cases, 558 executions, 0 failures passing
keystone-engage/agent-v1 engage baseline 100/100 (regression 70, architecture 25, edge 5) passing

The retrieval baseline reports access-control behavior as a first-class metric: all eight adversarial cases that attempt to retrieve out-of-scope records are blocked, and the system fails closed on 5 of 6 (83%) of the ambiguous cases in the retained 2026-04-11 baseline. The single miss (FC-005) has a demo-grade domain-scope guard merged 2026-05-17, with re-verification not yet recorded. Selected governance-related behaviors are included as explicit assertions in the evaluation.

Retained failing runs

An earlier 66-case pre-run (keystone-core/agent-v0-pre) passed; expanding the evaluation to 186 spec-compliant cases is what surfaced defects the smaller run had missed. That expanded run, keystone-core/agent-v0, produced nine failing cases traced to four distinct root-cause implementation defects. It is preserved as a retained artifact next to the passing keystone-core/agent-v1 baseline. The failing run is not an embarrassment. In this lineage, the retained failing run shows that the evaluation configuration detected specific implementation defects in the evaluated system. It is evidence about the evaluation method, not a proof of the system's substantive correctness, safety, or compliance.

This is a discipline the contact-center industry built long ago for compliance and quality management: bad calls were not hidden, they were analyzed. Keystone applies the same discipline to model behavior. See contact-center heritage for where this practice comes from.

The published ledger

The evaluation ledger is published at keystone-ledger. It holds the retained artifacts (case sets, run metadata, and raw results) that back every figure in the table above. The retained artifacts allow a reviewer to inspect the recorded cases, run metadata, and raw results associated with a published metric. Reproducing a historical execution additionally depends on reconstructing the relevant software, models, data, and environment.

The evaluation framework

keystone-verify is the current standalone HTTP evaluation harness for compatible endpoints. It is open source and targets those endpoints through evaluation profiles. The historical keystone-core/* artifacts listed above are retained in keystone-ledger and should not be assumed to have been produced by the current Verify CLI. The current framework provides:

  • A profile system (JSON) for declaring endpoint connection details and mapping response fields into the judge model.
  • A separate cases file (JSONL) containing request dictionaries and deterministic assertions.
  • A pure-function judge engine for severity, string conditions, minimum length, citation presence, fail-closed flags, and latency limits. It does not implement general semantic entailment or source-support verification.
  • A structured-artifact reporter that writes results.json and run_metadata.json to a selected local directory.
  • A CLI for running evaluations locally or in CI.

The reporter does not own historical retention or automatically publish output to keystone-ledger. Retaining both passing and failing runs is a separate repository and evaluation-process choice.

The keystone-verify page covers the full methodology and assertion model. Source: github.com/getkeystone/keystone-verify.

Historical naming

The ledger was formerly named keystone-kdat. Historical KDAT identifiers map to the current versioned names:

Historical Versioned
KDAT-001B keystone-core/retrieval-v1
KDAT-002C keystone-core/agent-v0
KDAT-002D keystone-core/agent-v1

Old references remain valid through this map; new baselines use the versioned convention only. These are historical artifact identifiers retained in keystone-ledger; keystone-core itself is not the public evidence repository.