SZL-Khipu-1.5B-abstain

<!-- SZL-ESTATE-CARD:v2:START --> <p align="center"><a href="https://a-11-oy.com/"><img src="https://huggingface.co/spaces/SZLHOLDINGS/README/resolve/main/assets/estate-banner-v2.svg" alt="SZL Holdings — governed, receipted, verifiable" width="100%"></a></p>

Control before action. Evidence after.

Product · Proof · GitHub · HF org

<!-- SZL-ESTATE-CARD:v2:END -->

EXPERIMENT. Adapter bytes present but unreceipted. Evaluators use SZLHOLDINGS/SZL-Khipu-1.5B until a receipt binds the bytes on this ID.

WEIGHTS ARE PRESENT ON THIS ID — card corrected 2026-09-25. Earlier versions of this card said "NO WEIGHTS IN THIS REPO". That is no longer true. Observed on 2026-09-25 (authenticated Hub tree read):

File	Bytes	SHA-256 (LFS)
khipu-abstain-adapter/adapter_model.safetensors	147,770,496	da0f948b7a6b555cbd50026eafcbb38ae8187b66be450fd1c3ed27f7eff9bef1
khipu-f16.gguf	3,093,668,832	0df16da8dc6d5865370b90880080dfefda028a5e4c0f758825bf40cdcfb93bd2

training_receipt.signed.json on this ID (Ed25519 keyId 89540347a69b789e, trainedAt 2026-09-08T18:44:18Z, finalTrainLoss 0.0250) binds adapterSha256 = bd5a1a92…9f1db and weightsArtifactSha256 = 86c33222…3476. Neither hash matches the bytes above. The receipt therefore does not attest these files. Until a receipt binds them, the weights are PRESENT_UNRECEIPTED: loadable, not evidence. khipu-abstain-adapter/README.md is an untouched PEFT template and carries no provenance.

What is receipted is the curriculum: train.jsonl, train.abstain.jsonl, adversarial.jsonl, eval.jsonl, khipu.schema.json (hashes in manifest.json and in the receipt match) and the 23 KB training script.

SZL-Khipu-1.5B-abstain

WEIGHTS: PRESENT_UNRECEIPTED. Bytes exist on this ID; no receipt binds them. Successor adapter with a receipted evaluation is SZLHOLDINGS/KHIPU-R2.

QLoRA adapter retrain recipe of the existing Khipu line. Raises in-memory ABSTAIN_OVERSAMPLE from 2 to 4 (8×4=32 abstain vs 15 navigate). Proposal-only. Λ = Conjecture 1. Doctrine v11 LOCKED 749/14/163.

<!-- SZL-ATELIER-CUT:v1:START -->

The cut

Most 'safety LoRAs' teach tone. This one teaches a binary: the handles are not enough. Research-only until abstain beats 2/6.

A specialist in silence. Capability is someone else's LoRA.

Silhouette → leave → SZL

No equivalent public combination was found among the leaders surveyed above (a snapshot, not an ecosystem-wide novelty claim).

Intended use

Stack on the navigator. Measure abstain. Do not ship on hope.

Limitations

• research-only
• proposal-only
• Does not magically fix 2/6 until a signed eval says so.
• The adapter bytes on this ID are not bound by any receipt; treat them as unverified.

Canonical GitHub: szl-holdings/szl-forge

<!-- SZL-ATELIER-CUT:v1:END -->

Evaluation

Status: NOT YET RUN. No fabricated k/n. publication_eligible is false until the held-out eval in train_khipu_abstain.py actually executes after training and its receipt binds the adapter hash above.

Prior original MEASURED abstain on SZLHOLDINGS/SZL-Khipu-1.5B is 2/6 (blocker). Eval protocol: eval.jsonl 5 navigate + adversarial.jsonl 6 abstain. Report k/n only.

Training (this job)

• Unsloth QLoRA, seed 11, lr 2e-4, adamw_8bit, train_on_responses_only, Trackio
• LoRA r=32 α=64, 45 epochs, ga=2, batch=1, constant_with_warmup (from train_khipu.py)
• Train: train.jsonl 15 navigate + train.abstain.jsonl 8 rows × 4
• Held-out never in gradients
• Script: train_khipu_abstain.py

Intended use

Proposal-only JSON retrieval plans (NAVIGATE / ABSTAIN) over synthetic Brain node handles. A controller outside the weights validates and resolves content. Not autonomous. Not a replacement for the signed original weights.

The estate