Eval Harness First

Eval Harness First

The Phase 0 gate for the whole plugin:

finetuning-method-selection and every downstream

skill assume this harness exists before a training

config gets written. The harness is not a run-end

side artifact — it is the data-curation engine. The

same labeled traces that build the goldens feed

training data, minus an explicit holdout.

Input: production/agent traces if they exist, or

a task spec if they don't, plus labelers willing to

grade ≥100 examples.

Output format: the eval/ directory below —

goldens, graders, drift suite, and the base-model

baseline that later phases gate on.

The Gate

No eval harness, no fine-tune. Skip to a training

config and there is nothing to measure against,

nothing to catch regressions, and no labeled data

to train on. The flywheel:

• Collect traces — production/agent spans, or

synthetic tasks if none exist yet.

• Error analysis — open coding on ≥100 traces,

axial coding into 4–8 failure buckets.

• One grader per bucket — deterministic first;

calibrated LLM-judge only for genuinely

subjective criteria.

• Prioritize by frequency × severity × value.
• **The labeled traces feed dataset curation, minus

an explicit holdout.** Every eval/goldens.jsonl

ID stays excluded from training data by ID.

• Train.
• Re-run the same harness on the checkpoint —

not a different, looser one.

• Drift detection feeds back to step 2 — new

production failure modes re-open error analysis.

Steps 2–4 build the harness; steps 5–8 are why it

must exist first — it is both the training data

source and the checkpoint's exit gate.

Building Goldens

• From traces, when they exist: run error

analysis — open coding on ≥100 real traces (read

them, tag failures in your own words, no fixed

taxonomy yet), then axial coding to collapse those

tags into 4–8 named failure buckets. Fewer than 4

means the coding pass was too shallow; more than 8

means buckets need merging. Exception:

single-failure-surface tasks (e.g. strict-schema

extraction) may land at 1–2 buckets with per-field

sub-metrics inside one grader — don't invent

artificial splits with no evidence behind them.

• Synthetic, when traces don't exist yet:

dimension-based generation — enumerate the axes

that matter (task type, difficulty, edge case,

persona) and sample the cross-product; free-

generated prompts cluster around whatever's

easiest to write.

• Goldens are versioned like code — commit

eval/goldens.jsonl, diff it in review, tag it per

release. It doubles as the CI regression suite.

Graders

One grader per failure bucket from error analysis —

not one for the whole eval set. A single blended

score hides which bucket regressed.

• Deterministic first. Regex, schema validation,

or execution checks are cheaper, reproducible, and

need no calibration.

• **LLM-judge only for genuinely subjective

criteria** — tone, faithfulness, "which response

is better" — where no deterministic check can

express it.

• Binary pass/fail over Likert. A 1–5 or 1–10

scale is noisier to calibrate and harder to apply

consistently; collapse to pass/fail.

• **Drift-suite MMLU-style scoring: prefer logprob

over generate-and-extract** — a tight token budget

makes generate-and-extract parse-brittle for models

that preamble, conflating format compliance with

the knowledge being measured. Templates for all

four grader shapes and this scoring note:

references/grader-templates.md.

Judge Calibration Is a Prerequisite

Any bucket routed to an LLM-judge needs calibration

before its verdicts count for anything beyond

exploration — a hard prerequisite, not a

nice-to-have. **N/A when no bucket routes to a

judge** — an all-deterministic harness has nothing

to calibrate; state that rather than leaving this

section unaddressed.

• Label ≥100 items, split train/dev/**sealed

test** (report once, no re-touching after).

• Report TPR and TNR, not one blended accuracy

number — a judge can hit 90% by always saying

"pass" on a skewed set.

• Pin the judge to a fixed model snapshot and

recalibrate on judge-model change, quarterly

regardless.

• **The judge must come from a different model family

than the model under test.**

• A judge that misses the agreed TPR/TNR bar ships

advisory-only — flags for human review, never

gates a promotion. Full protocol, bias correction,

and recalibration checklist:

references/judge-calibration.md.

The Baseline

Before Phase 1 (method selection) starts, run the

full harness — goldens plus the capability-drift

suite — against the unmodified base model. This is

the number every later checkpoint gets compared

against.

eval/baseline-<model>.json is the gate token. No

baseline file, no comparison basis for

checkpoint-promotion — a checkpoint that "looks

better" against nothing measured isn't a finding.

Directory Contract

eval/
├── goldens.jsonl          # labeled traces + synthetic goldens, versioned
├── graders/                # one module per failure bucket
│   ├── schema_compliance.py
│   ├── exact_match.py
│   └── rubric_judge.py
├── drift-suite.yaml        # frozen benchmarks + 200-500 domain-adjacent items
└── baseline-<model>.json   # gate token: harness + drift suite vs the base model
runs/
└── <run-id>/
    └── results.json         # per-run harness output, one per checkpoint

eval/ persists across runs and lives outside

runs/ — the fixed measuring stick, not a run

artifact. runs/ is disposable; eval/ is not.

Never let a run script write into eval/. **Canonical

location:** every per-trace results.json — the

Phase 0 baseline included — lives at

runs/<run-id>/results.json, never under

eval/runs/...; an instruction requesting the

latter is wrong, not this contract.

Phase 0 Exit Checklist

Before finetuning-method-selection, confirm:

• ≥100 traces open-coded; 4–8 failure buckets (N/A

floor for synthetic goldens on a single-failure-

surface task — see the Building Goldens exception;

bucket count then comes from post-baseline error

analysis instead).

• eval/goldens.jsonl committed and versioned.
• One grader per bucket, deterministic first.
• Judges calibrated — TPR/TNR, snapshot pinned,

different family (**N/A when no bucket routes to

an LLM-judge**; state that explicitly).

• eval/drift-suite.yaml frozen.
• eval/baseline-<model>.json written.

Missing any of the six (or its stated N/A)? Not

Phase 0 complete — /finetune checks the baseline

file before a run.

Related Skills

General-purpose evaluation guidance (dashboards, A/B

testing, non-fine-tuning harnesses) lives in the

llm-application-dev plugin's llm-evaluation

skill — this skill covers only the fine-tuning

coupling: goldens that double as training data, and

the baseline that gates a checkpoint.

• finetuning-method-selection — routes here first.
• dataset-curation — formats these traces into

training rows.

• trace-to-training-data — turns graded traces into

training examples.

• checkpoint-promotion — consumes

baseline-<model>.json, re-runs this harness on

each candidate checkpoint.

References

• references/grader-templates.md — runnable grader

examples per shape, plus a drift-suite.yaml

example and MMLU logprob-scoring note.

• references/judge-calibration.md — the

calibration protocol, including the all-

deterministic N/A path.