PostTrainLLM fine-tune report card · schema v1 · compiler 1.0.0

Qwen3-4B ReST Fused

— not recordedThe specialist package format does not record the owner goal that framed the run.

Target: Qwen3-4B ReST Fused
Base model: Qwen/Qwen3-4B-Instruct-2507 (bf16)
Candidate: qwen3-4b-rest-fused
Method: teacher-free ReST over checker-passing interleaved trajectories with a file-ops gold depth anchor
Compiled from: specialists/qwen3-4b-rest-fused (specialist-package)

Decision

Shipped Shipped for a named route only

retain only as a routed file-operations specialist; reject as a general successor and do not use as the Pace default planner

Routing constraint

retain only as a routed file-operations specialist; reject as a general successor and do not use as the Pace default planner. Do not use for: Pace default planner without re-distillation and its product ship gate; unqualified broad general-agent claims.

This candidate is safe only inside that envelope. It is not a general replacement for the base model.

Verification status

Not fully verified. This report does not claim a verified ship. Reasons:

Primary gate `file_ops_hard_gate` baseline is `historical`, not a current measurement.
Primary gate `file_ops_hard_gate` candidate is `historical`, not a current measurement.
Primary gate `file_ops_hard_gate` has no threshold value (state `missing`).
Primary gate `file_ops_hard_gate` has no passed value (state `missing`).
Frozen-eval identity is not recorded as a current measurement.
Train/eval overlap (leakage) was not checked with current evidence.

Failure reason: Not applicable
Failure-reason confidence: not-applicable
Lesson: Not recorded

Next action

— not recordedThe specialist package format records no machine-readable next action. The release action lives in the public artifact registry (docs/factory/public-artifacts.md) as prose.

Before and after

This candidate does not present an unconditional win: out_of_domain_breadth failed. Target and regression gates are reported independently below.

Every gate with baseline, candidate, derived delta, threshold, result, sample size, and frontier-ceiling evidence.
Gate	Role	Metric	Baseline	Candidate	Delta	Threshold	Result	n	Frontier ceiling
file_ops_hard_gate	primary	file_ops_hard_gate	0.7500 historicalRecorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).	1.0000 historicalRecorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).	0.2500 derivedDerived from at least one non-current value; inherits the weaker provenance of its inputs.	— not recordedThe specialist package format records no per-gate threshold.	— not recordedThe specialist package records no ship threshold for the primary gate, so a pass/fail result cannot be derived.	12 historicalRecorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).	1.0000 historicalRecorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).
out_of_domain_breadth	breadth	out_of_domain_breadth	0.6667 historicalRecorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).	0.5556 historicalRecorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).	-0.1111 derivedDerived from at least one non-current value; inherits the weaker provenance of its inputs.	— not recordedThe specialist package format records no per-gate threshold.	no derivedNo threshold was recorded. Derived as failing because the candidate scored below the baseline on a non-primary gate.	45 historicalRecorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).	0.9778 historicalRecorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).

Eval identity

Which suite produced each gate, how it was invoked, and whether it is frozen.
Gate	Suite	Command	Date	Frozen
file_ops_hard_gate	file_ops_hard_gate	— not recordedThe specialist package format records no eval command, so this gate cannot be replayed from the report card alone.	2026-09-03	— not recordedThe package does not record whether this suite is frozen.
out_of_domain_breadth	out_of_domain_breadth	— not recordedThe specialist package format records no eval command, so this gate cannot be replayed from the report card alone.	2026-09-03	— not recordedThe package does not record whether this suite is frozen.

Per-slice evidence

No slice evidence was recorded for this candidate.

Cost and performance

Latency, memory, throughput, and training cost/time. Absent evidence is reported as not measured, never as zero.
Metric	Value	Source
Latency	— not recordedTraining duration and normalized per-request latency were not measured. Eval time is candidate depth plus breadth wall time; throughput and peak RSS are the candidate depth values. Per-suite values and raw receipt hashes are preserved in the published result.	`specialists/qwen3-4b-rest-fused/eval_report.json#performance.latency_ms`
RAM / peak RSS	7755.0781 MB historicalRecorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).	`specialists/qwen3-4b-rest-fused/eval_report.json#performance.peak_rss_mb`
Decode throughput	13.3970 tok/s historicalRecorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).	`specialists/qwen3-4b-rest-fused/eval_report.json#performance.tokens_per_second`
Training time	— not recordedTraining duration and normalized per-request latency were not measured. Eval time is candidate depth plus breadth wall time; throughput and peak RSS are the candidate depth values. Per-suite values and raw receipt hashes are preserved in the published result.	`specialists/qwen3-4b-rest-fused/eval_report.json#performance.training_time_seconds`
Training cost	0 USD historicalTeacher-free local ReST iteration; no paid model API was used. Recorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).	`specialists/qwen3-4b-rest-fused/eval_report.json#performance.training_cost_usd`
Eval time	2011.1948 s historicalRecorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).	`specialists/qwen3-4b-rest-fused/eval_report.json#performance.eval_time_seconds`

Eval validity and leakage

Whether the benchmark is a trustworthy ruler: frontier ceiling, frozen-eval identity, and train/eval overlap.
Check	Result	Source
Frontier ceiling	1.0000 historicalRecorded evidence quality: fresh-paired-frontier-qualified-result-with-hashed-raw-receipts. Imported from a committed specialist package rather than a canonical factory-run folder, so it lacks current run provenance (command, hashes, raw predictions).	`specialists/qwen3-4b-rest-fused/eval_report.json#scores[0].frontier`
Frozen eval	— not recordedThe package records no frozen held-out split identity.	`specialists/qwen3-4b-rest-fused/eval_report.json`
Train/eval overlap	— not recordedNo train/eval overlap check is recorded for this package.	`specialists/qwen3-4b-rest-fused/eval_report.json`

Known eval limitations

Values come from a committed specialist package, not a canonical factory-run folder: eval commands, dataset hashes, and raw predictions are unavailable for independent replay.

Caveats

The fresh candidate won file-ops depth 12/12 versus 9/12 and removed all eight stock side effects.
The same candidate lost frontier-qualified breadth 25/45 versus stock 30/45, so the general-successor decision is reject.
The tracked result preserves raw receipt hashes; raw traces remain local and gitignored.
Pace requires a different intent envelope and product-specific ship gate.
Do not use for: Pace default planner without re-distillation and its product ship gate
Do not use for: unqualified broad general-agent claims

Source evidence

Every number above traces to one of these artifacts. Content hashes are recorded where the source file is committed.

model cardspecialists/qwen3-4b-rest-fused/model_card.md · committed-package-file · sha256 036dce9f2b45629b…
eval reportspecialists/qwen3-4b-rest-fused/eval_report.json · committed-package-file · sha256 e9924e33ad56792a…
reproducibility lockspecialists/qwen3-4b-rest-fused/tinygpt.lock.json · committed-package-file · sha256 9a699a938a99ac6f…
prompt contractspecialists/qwen3-4b-rest-fused/prompt.md · committed-package-file · sha256 1571ecc23b132713…
specialist registry entryspecialists/registry.json · committed-registry · sha256 34b7b43a62dcece3…
recorded result sourceevals/verified-wins/rest-requalification-result-v1.json · historical-record · sha256 368104c72ef06c66… · The document the legacy score was recorded in.
published weightshf://models/posttrainllm/qwen3-4b-rest-fused · external-artifact · Public weight location; not hashed by this compiler.

How to read the evidence states

measured: Read directly from a source artifact for this candidate.
derived: Computed from other recorded values (for example a delta).
historical: Imported from a legacy record without current canonical provenance. Treat as weaker than a measurement.
skipped: Deliberately not run for this candidate.
not recorded: Evidence should exist but does not. No number is implied.
not applicable: The check does not apply to this candidate.