SZLHOLDINGS/KHIPU-R2

<p align="center"> <img src="holo-banner.svg" alt="KHIPU-R2 - holographic step-up banner" width="100%"/> </p>

<h1 align="center">K H I P U - R 2</h1>

<p align="center"><em>Failure is an artifact, not a footnote.</em></p>

<p align="center"> <img alt="Base: Qwen2.5-1.5B-Instruct" src="https://img.shields.io/badge/base-Qwen2.5--1.5B--Instruct-334155?style=flat-square"/> <img alt="Downloads" src="https://img.shields.io/huggingface/dt/SZLHOLDINGS/KHIPU-R2?style=flat-square&color=3af4c8&label=downloads"/> <img alt="Artifact: QLoRA adapter r=32 a=64" src="https://img.shields.io/badge/artifact-QLoRA%20adapter%20r32%20a64-0d9488?style=flat-square"/> <img alt="Abstain: MEASURED 3/6, was 2/6 - not a pass" src="https://img.shields.io/badge/abstain-3%2F6%20was%202%2F6%20%C2%B7%20not%20a%20pass-b45309?style=flat-square"/> <img alt="Plan-valid 11/11" src="https://img.shields.io/badge/plan--valid-11%2F11-16a34a?style=flat-square"/> <img alt="Hallucinated citations: 0" src="https://img.shields.io/badge/hallucinated%20citations-0-16a34a?style=flat-square"/> <img alt="Re-verified 2026-09-13" src="https://img.shields.io/badge/re--verified-2026--09--13-16a34a?style=flat-square"/> <img alt="Lifecycle: research-only" src="https://img.shields.io/badge/lifecycle-research--only-7e8aa3?style=flat-square"/> </p>

<p align="center"> <code>KANCHAY</code> - Doctrine v11 - Lean <code>749/14/163</code> - Conjecture 1 (advisory) - <a href="https://a-11-oy.com">a-11-oy.com</a> </p>

Adapters are on this repo. Abstain is MEASURED 3/6, not a pass. This card-only release does not modify adapter files and does not open the publication or autonomy gates.

QLoRA adapter on disclosed Qwen/Qwen2.5-1.5B-Instruct (runtime unsloth/Qwen2.5-1.5B-Instruct-bnb-4bit). Proposal-only brain navigator / abstain retrain. Doctrine v11 LOCKED. Conjecture 1 (advisory, never a theorem).

Specification

Evaluation (MEASURED this job)

Method: in-process Unsloth generate, scoring ported from eval_khipu.py, temperature 0, held-out never in gradients. Host job worker. Date 2026-08-28 17:20 UTC. File: eval_measured.json.

Prior published original SZL-Khipu-1.5B MEASURED abstain was 2/6. This run is 3/6. Small n. Do not derive a world-rank score from k/n on n=11.

Independent re-verification (2026-09-13)

Owner-host re-run with a repaired szl_khipu_eval.py. The harness had a decoding bug (tok.decode on a 2-D batch tensor inside generate_plan) that returned prompt characters instead of model output; every prior local fp16 score produced by the buggy harness is void. Three CPU fp16 evaluations, temperature 0, held-out set unchanged:

All runs reproduce the 2026-08-28 job receipt exactly. File: eval_verified_2026-09-13.json. GPU UNAVAILABLE (CPU eval). This is verification evidence, not promotion; publication and autonomy gates remain closed.

Training (MEASURED / REPORTED)

• —Unsloth QLoRA, seed 11, lr 2e-4, LoRA r=32 a=64, 45 epochs
• —Train: 15 navigate + 8 abstain rows x oversample 4 (in-memory 32)
• —Held-out: 5 + 6, held_out_in_gradients: false
• —training_loss MEASURED 0.017188 is a train metric, not an eval
• —adapter sha256 e44d53f29f2d443598e06d6c0441557fd3a5010888c7aa97b56ec3c0e050d349

What this is NOT

• —Not a replacement for SZL-Khipu-1.5B
• —Not Chaski (Qwen3.5 lock)
• —Not an autonomous agent
• —Not a GGUF. Mini GGUFs exist on A11OY-MINI; Mini evals none-this-run; Mini does not inherit this 3/6

Load

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base_id = "Qwen/Qwen2.5-1.5B-Instruct"
tok = AutoTokenizer.from_pretrained(base_id)
base = AutoModelForCausalLM.from_pretrained(base_id)
model = PeftModel.from_pretrained(base, "SZLHOLDINGS/KHIPU-R2")

Owner: Stephen Lutar / SZL Holdings.