README.md

---
license: apache-2.0
base_model: Qwen/Qwen2.5-Coder-7B-Instruct
tags:
  - uncensored
  - abliterated
  - capability-preserving
  - certified
  - ektome
  - sphragis
  - qwen2.5
language:
  - en
pipeline_tag: text-generation
---

![Ektome-Qwen2.5-Coder-7B-Instruct-PristinelyUncensored](./hero.png)

# Ektome-Qwen2.5-Coder-7B-Instruct-PristinelyUncensored

**Uncensored — and carrying a statistical certificate that it wasn't damaged.**

> **Capability retention certified against the pristine model at n=2800.**

$$\colorbox{black}{$\color{white}
\begin{array}{ll}
\textsf{EKTOME CERTIFICATE} & {} \\
\textsf{capability} & \textsf{PASS} \\
\textsf{margin} & 3\% \\
\textsf{items } n & 2800 \\
\textsf{worst-axis bound} & +0.010 \\
\textsf{compliance} & \textsf{not recorded} \\
\end{array}$}$$


📄 **[Read the whitepaper (PDF)](./whitepaper.pdf)** — full method, receipts and certification.
The PDF is the authoritative document: dark-typeset, with the complete derivation, the
per-axis certificate and the reproducibility hashes.

---

## Why this exists

Standard abliteration removes a coarse *refusal direction* that is entangled with
directions carrying knowledge and reasoning. The result is an uncensored model with a
capability tax that is **almost never measured**.

Ektomē (ἐκτομή, *excision*) isolates and removes only the refusal-**specific**
component, leaving general helpfulness intact, and does so norm-preservingly on the
pristine model — no training, no distillation, no damage to repair. The extraction
depth is selected per model by automated search against measured compliance.

The estimator, excision operator and depth-selection procedure are proprietary.
What is published here is the **measured outcome** and the evidence for it, which you
can verify against the artifacts in this repo.

## The receipt

_No compliance/MMLU receipt was recorded for this model. The evidence below is the certificate._


## The certificate

Capability retention is certified by a paired non-inferiority test against the pristine
model (exact McNemar, Holm-corrected, one-sided bootstrap bound on the drop $d$ vs a
3% margin):

| axis | n | ref | cand | d upper | verdict |
|---|---|---|---|---|---|
| arithmetic | 1400 | 0.864 | 0.865 | +0.003 | PASS |
| instruction | 600 | 0.663 | 0.660 | +0.010 | PASS |
| knowledge | 400 | 0.968 | 0.968 | +0.000 | PASS |
| reasoning | 400 | 0.953 | 0.953 | +0.000 | PASS |


**Overall: PASS (3% margin, n=2800, alpha=0.05)**

Reproducible from `seed=20260726`, pack `sha256:7bbaff877146e081…`.

### Generation health checks

_Not recorded for this model._


## Quantisations

| file | bits | notes |
|---|---|---|
| `Ektome-Qwen2.5-Coder-7B-Instruct-Q8_0.gguf` | 8 | near-lossless |
| `Ektome-Qwen2.5-Coder-7B-Instruct-Q6_K.gguf` | 6 |  |
| `Ektome-Qwen2.5-Coder-7B-Instruct-Q5_K_M.gguf` | 5 |  |
| `Ektome-Qwen2.5-Coder-7B-Instruct-Q4_K_M.gguf` | 4 | imatrix |
| `Ektome-Qwen2.5-Coder-7B-Instruct-IQ4_XS.gguf` | 4 | imatrix, smallest usable |
| `Ektome-Qwen2.5-Coder-7B-Instruct-IQ3_M.gguf` | 3 | imatrix |

`IQ*` variants are imatrix-quantised — better quality per bit at low precision.


## Limitations

The certificate bounds **capability retention only**. It does not certify safety, factual
accuracy, or fitness for any purpose. Axes marked *inconclusive* are honestly
under-powered, and the certificate states the $n$ needed to resolve them. Compliance uses
a keyword classifier — a proxy that evasive phrasing can fool. **This model is uncensored
by construction: it will not refuse, and you are accountable for what you do with it.**

## Citation

```bibtex
@software{ektome_Ektome-Qwen2.5-Coder-7B-Instruct-PristinelyUncensored,
  title  = {Ektome-Qwen2.5-Coder-7B-Instruct-PristinelyUncensored},
  author = {Zynerji},
  year   = {2026},
  url    = {https://huggingface.co/Zynerji/Ektome-Qwen2.5-Coder-7B-Instruct-PristinelyUncensored}
}
```