	pristine	Ektomē
compliance (n=100, AdvBench)	0.000	0.990
MMLU-val (n=400, 0-shot loglik)	0.8625	0.855 (dcap −0.0075, SE 0.017)
foreign / degen / instr	0.000 / 0.000 / 0.400	0.000 / 0.000 / 0.400
vision	yes	yes — verified by generation

Winner: single rank-1 direction at layer_frac=0.65, 128 residual-write matrices (16 full-attention o_proj + 48 linear-attention out_proj + 64 MLP down_proj).

Two defects in the previous upload, both fixed

1. The vision tower was missing. Qwen/Qwen3.6-27B is Qwen3_5ForConditionalGeneration — natively multimodal. The previous artefact was Qwen3_5ForCausalLM with no vision_config: the loader tried AutoModelForCausalLM first and fell back to multimodal classes only on an exception. For qwen3_5 that call does not raise — it returns the text sub-model — so the fallback never fired and the tower was dropped without a warning. The loader now selects the class from config.architectures up front. Vision in this build is confirmed by asking the model to describe a synthetic image, not by the presence of a config key.

2. The reported capability baseline was wrong. The previous card reported MMLU 0.610 → 0.598. That 0.610 was measured on the mis-loaded text-only model. On the same 0-shot log-likelihood harness, the correctly-loaded base scores 0.8625 — reproduced three times independently. The old figure was attributed to a harness difference; it was not, it was the loader. The old and new MMLU pairs are therefore not comparable, and the improvement claim implied by putting them side by side would be false. What is comparable is the lever: the previous build needed per-layer directions plus over-ablation at weight 1.4 to reach full compliance; this one reaches 0.990 with the mildest rung on the ladder.

Additionally, the refusal direction is now estimated with the chat template rendered in the same mode the behaviour is scored in (enable_thinking=False). Qwen3.6-27B is a thinking model, and the two renderings end on different tokens, so the previous direction was estimated at "about to think" while refusal was scored at "about to answer".

⚠ The legacy `-Q4_K_M.gguf` in this repo is the OLD build

Ektome-Qwen3.6-27B-PristinelyUncensored-Q4_K_M.gguf (16.55 GB) dates from 2026-07-02. It is kept for continuity — it has real download history — but it is superseded on both counts described above:

it was converted from the text-only load, so it has no vision tower and no mmproj;
it was abliterated with the mis-templated refusal direction.

For a current GGUF use the -GGUF repo, and take the mmproj-*.gguf with it.

Quantisations

-NF4 — bitsandbytes 4-bit, ~17 GB, vision retained, compliance 1.000, MMLU 0.853
-GGUF — Q4_K_M / IQ4_XS / Q3_K_M plus mmproj-*.gguf (required for vision)

GPTQ is not possible on this architecture. The Marlin kernel requires out_features divisible by 64; Gated DeltaNet's linear_attn.in_proj_a / in_proj_b emit one scalar per value head — 48 here. gptqmodel skips those layers during packing but still lists them in the quantisation manifest, producing a checkpoint that loads uninitialised weights and fails with CUBLAS_STATUS_INTERNAL_ERROR. Use NF4 or GGUF.

Method

Ektomē: a rank-1 projected (orthogonalised against the harmless mean), norm-preserving excision of the refusal direction from residual-write matrices (o_proj, out_proj, down_proj). Zero training, zero fine-tuning, zero distillation — the delta is a linear weight edit derived from activations. A configuration ships only if compliance clears target and capability holds (dcap ≥ −SE) and generation quality holds (no code-switching, degeneration, or instruction-following loss).

Caveats

Compliance uses an opening-anchored keyword classifier — a proxy that evasive phrasing can fool. Validated by a pristine control: the un-edited base reads ~100% refusal.
MMLU here is 0-shot log-likelihood at n=400, so SE ≈ 0.017–0.020. Differences smaller than that are noise, and are reported as such rather than as improvements.
The certificate bounds capability retention. It does not certify safety, factual accuracy, or fitness for any purpose. This model is uncensored by construction: it will not refuse.

Zynerji/Ektome-Qwen3.6-27B-PristinelyUncensored

Ektome-Qwen3.6-27B-PristinelyUncensored

Two defects in the previous upload, both fixed

⚠ The legacy -Q4_K_M.gguf in this repo is the OLD build

Quantisations

Method

Caveats

Zynerji/
Ektome-Qwen3.6-27B-PristinelyUncensored

⚠ The legacy `-Q4_K_M.gguf` in this repo is the OLD build