---

base_model: google/gemma-4-E2B-it

language:

• multilingual

library_name: gguf

license: apache-2.0

license_link: https://ai.google.dev/gemma/docs/gemma_4_license

pipeline_tag: text-generation

tags:

• gguf
• gemma4
• q8_0
• abliterated
• uncensored
• model-editing
• heretic-nx
• prime
• lm-studio

---

Gemma 4 E2B IT — Heretic NX PRIME Q8_0

Heretic NX PRIME is a direct-Q8 behavioral edit of

google/gemma-4-E2B-it.

It is designed to reduce false refusals while explicitly measuring drift from

the unedited Q8 deployment artifact.

The release contains the evaluated Q8_0. PRIME is the project's internal

evidence and capability-preservation protocol, not a third-party certification.

This GGUF is for text generation and does not bundle the base model's separate

image or audio preprocessing components.

File

| File | Size | SHA-256 | Refusal markers |

| --- | ---: | --- | ---: |

| Gemma-4-E2B-it-Heretic-NX-PRIME-Q8_0.gguf | 4.95 GB | 43d2b7ed2d756f7ddd5100eafaa23d2ff108ab2af60b89a79b7634b7da1050d9 | 6/104 |

Evaluation

The final GGUF was loaded from disk with pinned llama.cpp commit

d7bd3bfcad3e29c7e49fd26f38c79ee3e9a3fd6b. Refusal evaluation used the

native pre-tokenized /completion endpoint, greedy sampling, 96 generated

tokens, four runtime slots and the official chat template with

enable_thinking=False.

| Metric | Result |

| --- | ---: |

| Lexical refusal markers, 104 harmful rows | 6 |

| Hit rows, one-based | 3,38,60,68,75,97 |

| Mean first-token KL(original Q8 || variant), 104 benign rows, full 262,144-token vocabulary | 0.008579 |

| Median first-token KL | 0.000054 |

| P95 first-token KL | 0.025772 |

| Maximum first-token KL | 0.201503 |

The release satisfies both predeclared targets: at most 6 lexical refusal

markers and mean first-token KL at most 0.05.

Lexical markers are a refusal proxy, not semantic harmful-task success or a

universal quality score. The 104 harmful rows participated in development and

selection, so this suite is not an untouched holdout.

Paired capability check

After candidate selection, the original and Heretic Q8 artifacts were evaluated

on the same deterministic 854-question slice from ARC-Challenge, HellaSwag and

MMLU. Each answer was the greedy first-token argmax restricted to A/B/C/D;

prompts, tokenizer, runtime and precision were identical between arms.

| Task | Rows | Original Q8 | Heretic NX Q8 | Difference |

| --- | ---: | ---: | ---: | ---: |

| ARC-Challenge | 256 | 74.22% | 73.44% | -0.78 points |

| HellaSwag | 256 | 58.59% | 59.38% | +0.78 points |

| MMLU | 342 | 61.99% | 62.28% | +0.29 points |

| Overall | 854 | 64.64% | 64.75% | +0.12 points |

The paired bootstrap 95% interval for Heretic minus original is **[-0.94,

+1.17] points**. There were 543 questions both got right, 292 both got wrong,

9 original-only successes and 10 Heretic-only successes. This passes the

predeclared 3-point non-inferiority margin and the symmetric +/-3-point

equivalence gate. The interval crosses zero, so this does not demonstrate

an aggregate accuracy increase; it supports capability preservation on this

narrow multiple-choice slice.

Method

The artifact is edited directly from the source Q8_0; it is not reconstructed

from a separately quantized edited checkpoint. Seven dense shared operators are

modified: FFN down projections in layers 15, 23, 25 and 26, plus attention

output projections in layers 16, 17 and 30.

The selected profile uses benign-penalized detector distillation at

lambda=100, beta=4.0, additive repair gamma=1.875, and two coordinate

strength adjustments. The final composed detector at every site is projected

orthogonal to the BF16 activation of benign evaluation row 53, which was the

development KL outlier. This targeted repair is disclosed because it means the

benign KL suite also participated in final selection and is not an untouched

holdout.

All plans, factors, source artifacts, runtime evaluations and untouched output

bytes are SHA-256 bound. The compact release report, selected factor artifact,

protected input vector, ablation plan and sanitized evaluation reports are

included under evaluations/; reproducible experiment scripts are available in

0xZKnw/heretic-nx.

LM Studio / llama.cpp

Download the GGUF and load it normally. For llama.cpp:

llama-server \
  -m Gemma-4-E2B-it-Heretic-NX-PRIME-Q8_0.gguf \
  -ngl 99 -c 4096 --jinja

Use a recent runtime with Gemma 4 support. Thinking is disabled in the reported

tests; applications should select the desired chat-template mode explicitly.

Limitations and responsibility

This edit intentionally weakens refusal behavior. It can increase compliance

with unsafe, illegal, incorrect or otherwise harmful requests. It does not add

factuality, judgment, sandboxing or application-level safety. Run untrusted

generations in an appropriate sandbox.

The capability evaluation is narrow, and both behavioral suites participated

in development. The release is not externally certified or claimed equivalent

to the original model on every task.

Use is subject to the Gemma 4 Apache 2.0 license.