← back to catalog · registered 2026-09-27 20:57

willmargs/Qwen3.8-27B-abliterated-arithmetic-tradeoff

willmargs Qwen 27B multimodal
curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/willmargs%2FQwen3.8-27B-abliterated-arithmetic-tradeoff"
Response includes
  • classification unknown
  • files 33
  • author_summary 1 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
?
Primary method

Unclassified

No clear signals of an abliteration technique in this model.
Confidence
UNKNOWN
Why this label 1 signal
No classification signals present. This may not be an abliterated model at all - it could be a repackaging, a merge with unrelated goals, or unrelated content that mentions the term.
  • no classification signals present (no abliterated, uncensored, or known producer/method markers)
Refusal direction extraction

No specific extraction method could be identified for this model. The producer either did not document it or used a proprietary pipeline.

What is a refusal direction? →
Downloads · 30-day
3
Likes
0
Model age
3d ago
created 2026-09-24

Genealogy 0 direct forks

Full fork graph →

This model's place in the market. Above: what it was derived from. Below: the tree of everything derived from it.

Metadata

License
apache-2.0
Tags
transformers safetensors qwen3_5 image-text-to-text abliterated research refusal-lab conversational arxiv:2406.11717 base_model:Qwen/Qwen3.8-27B base_model:finetune:Qwen/Qwen3.8-27B license:apache-2.0

Related

Total size
51.7 GB
Files
33
Quantizations
1
Registered
2026-09-27 20:57
Last updated on HF
2026-09-27 17:41

Files by quantization

Auxiliary files 33 files 51.8 GB
model-00004-of-00018.safetensors 3.72 GB 146a1877 download
model-00016-of-00018.safetensors 3.71 GB b817e899 download
model-00006-of-00018.safetensors 3.71 GB 63cdafea download
model-00008-of-00018.safetensors 3.71 GB 3c659dee download
model-00010-of-00018.safetensors 3.71 GB 45f0e8f0 download
model-00012-of-00018.safetensors 3.71 GB baa47985 download
model-00014-of-00018.safetensors 3.71 GB ed8da81f download
model-00001-of-00018.safetensors 3.69 GB 646fa080 download
model-00018-of-00018.safetensors 3.16 GB 1d347950 download
model-00002-of-00018.safetensors 2.83 GB fe0281be download
model-00003-of-00018.safetensors 2.37 GB 989bbfdf download
model-00007-of-00018.safetensors 1.96 GB 5b992406 download
model-00009-of-00018.safetensors 1.96 GB 4971a170 download
model-00011-of-00018.safetensors 1.96 GB 85d8491e download
model-00013-of-00018.safetensors 1.96 GB 73c40a10 download
model-00015-of-00018.safetensors 1.96 GB f61484da download
model-00017-of-00018.safetensors 1.96 GB 142a2ad2 download
model-00005-of-00018.safetensors 1.96 GB 2a1a946f download
tokenizer.json 12.2 MB 0997f410 download
vocab.json 6.41 MB 0aa0ce06 download
merges.txt 3.20 MB a494e019 download
model.safetensors.index.json 110 KB da35e3c5 download
BASE_MODEL_README.md 63.5 KB bc8aa0e3 download
upload_manifest.json 38.7 KB 814be728 download
tokenizer_config.json 17.5 KB 5de744b3 download
LICENSE 11.3 KB f938136e download
chat_template.jinja 8.74 KB c0c686f9 download
README.md 4.23 KB bfb03e10 download
config.json 4.21 KB 706cebd7 download
.gitattributes 1.53 KB 52373fe2 download
preprocessor_config.json 390 B 2ea84a43 download
video_preprocessor_config.json 385 B 3ba673a5 download
generation_config.json 202 B 023756cf download

README current version from Hugging Face


license: apache-2.0
base_model: Qwen/Qwen3.8-27B
base_model_relation: finetune
library_name: transformers
pipeline_tag: image-text-to-text
tags:

  • abliterated
  • research
  • refusal-lab
  • qwen3_5

This description is entirely AI generated and largely unreviewed by me

Qwen3.8-27B Abliterated — BF16

A BF16 variant of Qwen3.8-27B with a refusal-associated rank-one weight edit.
It shows higher aggregate multiple-choice arithmetic accuracy than the base
model on the reported evaluation, alongside reduced refusals on an adult-fiction
willingness probe.

  • Parameters: 27.8B
  • Format: BF16 safetensors, 18 shards, approximately 55.6 GB; not GGUF/4-bit
  • Base: Qwen/Qwen3.8-27B, revision 1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0
  • License: Apache 2.0
  • Evaluated mode: text input, thinking disabled

The weight edit is already applied. No steering hooks, prompt router, or external
calculator are required. Do not apply the accompanying direction again; load the
original Qwen checkpoint if you want the unmodified model.

Evaluation

Measure Base model This weight edit
Multiple-choice arithmetic correct 189/252 (75.0%) 197/252 (78.2%)
Numeric-answer arithmetic correct 63/63 (100%) 63/63 (100%)
Adult-fiction willingness refusals 64/64 0/64

The MCQ result is a net improvement of eight correct answers: 15 base-model
mistakes were fixed and seven previously correct answers became incorrect.
The numeric-answer set uses different questions from the multiple-choice set.

Measurements characterize this BF16 weight-edit configuration. Uploaded files
were integrity-checked, but no separate inference benchmark was run after upload.
Detailed evaluation data are available in the report
and the evaluation summary.

Evaluation limits

  • The refusal probe asks for a brief willingness statement, not the requested
    content
    . It does not establish full-completion NSFW behavior.
  • Eleven benign willingness responses were unclassified or truncated in the
    tested format. Aggregate arithmetic improvement does not establish preservation
    of every answer or of broader capabilities.
  • Thinking-enabled reasoning, long contexts, vision, quantized versions, and
    broad capability/safety benchmarks have not been evaluated for this edit.

Loading

Use sufficient GPU memory for the BF16 checkpoint or supported device offloading.
The evaluation runtime used Torch 2.10.0 and Transformers 5.17.0, with
flash-linear-attention 0.5.2 for fast linear-attention kernels.

import os
import torch
from transformers import AutoTokenizer, AutoModelForImageTextToText

repo = "willmargs/Qwen3.8-27B-abliterated-arithmetic-tradeoff"
# Supply HF_TOKEN securely in the process environment for private-repo access.
token = os.environ["HF_TOKEN"]
tokenizer = AutoTokenizer.from_pretrained(repo, token=token)
model = AutoModelForImageTextToText.from_pretrained(
    repo, token=token, dtype=torch.bfloat16, device_map="auto"
).eval()
text = tokenizer.apply_chat_template(
    [{"role": "user", "content": "Calculate 17 + 24. Reply only with the number."}],
    tokenize=False, add_generation_prompt=True, enable_thinking=False,
)
inputs = tokenizer(text, return_tensors="pt", add_special_tokens=False).to(model.device)
with torch.inference_mode():
    output = model.generate(**inputs, do_sample=False, max_new_tokens=32)
print(tokenizer.decode(output[0, inputs.input_ids.shape[1]:], skip_special_tokens=True))

Quantizing these weights is a separate transformation and may change behavior.

Model details and attribution

The edit projects a shared direction out of 129 text residual-write matrices
(input embeddings, attention outputs, and MLP down-projections). The original
architecture, tokenizer, vision weights, and license are retained. This does not
establish a unique refusal neuron or mechanism.

Base model: Qwen/Qwen3.8-27B.
Original model card: BASE_MODEL_README.md.
License: LICENSE. Modified weights published by willmargs.
Method background: Refusal in Language Models Is Mediated by a Single Direction.

Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.