checkpoint-promotion skill

Checkpoint Promotion

The Phase 5 gate for the whole plugin: a checkpoint that trains cleanly and beats its task metric still doesn't ship without clearing all four stages below. eval-harness-first built the suite re-run here — this skill is where that suite's baseline decides something.

Input: a trained checkpoint, eval/baseline-.json from eval-harness-first, and the frozen eval/drift-suite.yaml. Output format: promotion-report.md — the four-stage evidence plus a terminal PROMOTE or REJECT verdict that /finetune Phase 5 and /promote-checkpoint consume directly.

The Four-Stage Gate

Each stage gates the next — a failure at stage 2 means stage 3 doesn't run. Stages 2 and 3 share one expensive inference pass, so running them concurrently and applying gate order at verdict time is licensed on a deterministic arena (nothing saved by serializing); a judge-based arena should still wait for stage 2 first — that's where the real savings are.

any eval touches the checkpoint: dedup the training set, check for eval-goldens leakage (the exact failure trace-to-training-data's Hygiene section exists to prevent), and scan for label noise. A checkpoint trained on leaked goldens invalidates every later stage.

1. Data-quality gate. Before

capability-drift suite.** Re-run eval-harness-first's eval/drift-suite.yaml — MMLU/GSM8K/IFEval plus 200–500 domain-adjacent items — against the checkpoint and diff against baseline-.json per benchmark against the Drift Budget table below.

1. **Held-out + frozen

Position-randomized judge, checkpoint vs. base model, same prompts — or the deterministic paired-comparison variant in references/gate-templates.md when every grader in the harness is deterministic (no LLM-judge; position randomization N/A there). A holdout win that loses the live arena does not ship — stage-2 numbers and stage-3 judgments must agree; a win on frozen goldens and a loss in paired comparison is a real signal, not a discrepancy to explain away.

1. Paired arena vs. base.

rollout with auto-rollback for any checkpoint reaching production traffic. Local-only users stop at stage 3 — skipping stage 4 for a local deployment is the correct stopping point, not a shortcut.

1. Canary. 5–10% stratified

Drift Budget

The >5pt row governs regardless of the others: a checkpoint that gained 8 points on the target task and lost 6 points of general capability still fails here — task improvement never buys back a drift-budget breach.

Item count derives from the budget, not convenience: the strict n for a half-width under half the 5pt hard-fail threshold is ~1,300 at typical accuracy (p≈0.7); n=200 is a pragmatic floor (±6pt half-width at that same p, n=50 ±13pt) — report the half-width with every verdict, and treat a margin smaller than it as REJECT (uncertain), not PASS/HARD FAIL. Full math and a 5-run cautionary example: references/gate-templates.md.

RERUN is not a verdict. A 2–5pt drift only ever produces a PROMOTE or REJECT after the seed-variation rerun completes — PROMOTE requires landing back at ≤1pt (noise); any rerun still

1pt — 2–5pt band or >5pt breach

alike — resolves stage 2 to a hard REJECT. No report may reach the Verdict section with stage 2 still showing RERUN.

Catastrophic Forgetting

Unmanaged LoRA fine-tuning loses real general capability, and stage 2 is what catches it:

unmanaged** — no replay, no regularization.

• **~43% knowledge loss

— some replay or a conservative LR.

• ~10% with basic management

disciplined case.

• ~3% with replay + EWC — the

mix is the standard mitigation** — blend general- domain data into training rather than target-task data alone.

• **10–30% general-data replay

If a checkpoint hits the >5pt hard fail in stage 2, work this escalation ladder in order — the one canonical order this skill and references/gate-templates.md both point to:

fraction — swap rows, don't add them** (adding confounds fraction with total optimizer steps). Dose is not monotonic at small-run scale (<~100 steps) — re-check drift after any swap.

1. **Adjust the replay-mix

same rank/LR levers lora-qlora-recipes and preference-optimization tune for the training run, applied here in reverse.

1. Lower the learning rate.
2. Fewer epochs.
3. A smaller LoRA rank — the

This order is a default, not a law: remediation guidance from a single before/after run pair is a hypothesis — label it low-confidence once any lever produces a reversal, and prefer a seed-variation repeat over trusting the next rung blindly. A lever that clears the drift breach but drops a success-criterion metric below target is a two-sided tradeoff for a human, not a reason to keep descending the ladder. Full reasoning and the 5-run trajectory behind both caveats: references/gate-templates.md.

Disclose drift-suite instruction reuse. A replay row copying the drift harness's exact instruction phrasing (not just disjoint source items) makes that benchmark's post-replay score an upper bound — flag it instruction-familiar, or re-probe with a paraphrase, before treating a near-budget pass as clean.

The Verdict

promotion-report.md covers all four stages as sections and must end with a terminal verdict: PROMOTE or REJECT, the evidence that produced it, and exactly one top remediation when the verdict is REJECT. Template: references/gate-templates.md. The terminal contract other skills parse:

## Verdict

REJECT

Evidence: domain-adjacent drift
suite dropped 6.2pt (threshold:
>5pt hard fail) despite +8pt on
the target task.

Top remediation: swap the
replay-mix fraction from 10%
toward 20%, holding step count
constant.

error.** A checkpoint that fails stage 2's drift budget or stage 3's arena comparison did its job. Don't treat a REJECT as a failed run needing a rerun of this skill; it's the correct output of a working gate.

• **REJECT is a result, not an

menu.** Evidence sections may list everything observed; the verdict section names the single highest-leverage fix per the escalation ladder above. A report that hedges across three possible fixes hasn't done the prioritization this skill exists to do.

• **One remediation, not a

skill produces a verdict and a report, not a re-triggered training run. A REJECT hands the remediation back to a human decision at finetuning-method-selection or the relevant training skill.

• No auto-retraining. This

Related Skills

drift suite and baseline this skill re-runs and diffs against; no baseline-.json means nothing to gate against.

• eval-harness-first — owns the

valid next step after a PROMOTE verdict.

• quantized-export — the only

lora-qlora-recipes — own the LR and rank levers in the Catastrophic Forgetting escalation path; this skill diagnoses the breach, those skills own the config that caused it.

• preference-optimization and

replay-mix construction recipe the escalation ladder's first rung applies.

• dataset-curation — owns the

Complete promotion-report.md template with all four stages, the drift-suite scoring table, the paired-arena protocol (item count, position randomization, win-rate threshold), and a replay-mix configuration example: references/gate-templates.md.