Trained Models & Weights
Collection
Published model weights, quantizations, recipes, and their training/inference labs. Published means the repository exists; signed receipts and model-c β’ 11 items β’ Updated β’ 1
ReceiptAgent-Nano
ALLOW Β· DENY Β· ABSTAIN Β· ESCALATE. Escalation is a class, not a retry loop.
Family. nano Β· Evidence. SYNTHETIC Β· Weights. numpy Β· Params. 4-10-4
Hub: SZLHOLDINGS/ReceiptAgent-Nano
The cut
Anthropic refuses. NVIDIA rails. Unsloth trains. We add a fourth way: hand the decision to a human with a receipt. Loop-tax lives here.
A policy head that cannot silently succeed. Every output is one of four named gates.
Silhouette β leave β SZL
| Leader | Take, then tweak |
|---|---|
| Anthropic | Constitutional refuse β typed DENY/ABSTAIN. |
| NVIDIA | NeMo Guardrails flow β four-way head. |
| Unsloth | Grown form is SZL-Forge-1.5B-ReceiptAgent. |
Nobody else ships this combination. That is the point of a one-of-one.
Intended use
Fail-closed unit tests for the 4-way gate.
Limitations
- Synthetic 4-D features. Not a substitute for the 1.5B agent.
- Hub kernel labels are ALLOW/WARN/BLOCKED/ESCALATE β kernel is truth. This atelier MLP is a 4-class silhouette, not rule_check.
Honesty
| Claim | Label |
|---|---|
| This card's numbers | SYNTHETIC |
| Energy / joules | UNAVAILABLE unless a signed meter says MEASURED |
| Ξ uniqueness | Conjecture 1 OPEN β not a theorem |
| GGUF as the signed object | FALSE |
Doctrine v11 LOCKED Β· 749 declarations Β· 14 axioms Β· 163 sorries Β· locked-proven 8.
Apache-2.0. Copyright 2026 SZL Holdings Β· Stephen P. Lutar Jr. Β· ORCID 0009-0001-0110-4173.
Inference Providers NEW
This model isn't deployed by any Inference Provider. π Ask for provider support