← back to catalog · registered 2026-08-22 13:56

mckerm1t/gemma-4-e4b-it-abliterated-bf16

mckerm1t Gemma 7.5B
Your rig guess connected
? Why do I need an app?
Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model.

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/mckerm1t%2Fgemma-4-e4b-it-abliterated-bf16"
Response includes
  • classification m1
  • files 9
  • benchmarks 11 entries
  • hub_downloads_all_time 3,827
  • author_summary 1 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M1
Primary method

Direct removal

No other method signals detected in this model.
Confidence
MEDIUM
Why this label 3 signals
Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.
  • 'abliterated' in name/tags
  • is_gguf=0 (base model)
  • no specific method indicators - defaulting to M1 (most common)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
Downloads · lifetime
4K
309 last 30d - cooling
Likes
2
Model age
6mo ago
created 2026-04-11
Downloads over time
Now4K→from790↑400%
6321.8K3.1K4.3K790 on Apr 154K on Oct 11AprMayJunJulAugSepOct
Apr 15 → Oct 11 · 65 snapshots · spans 179 days

Benchmarks

Portrait before abliteration
Benchmarks of the base model as it stood before the refusal-removal operation. Compare with the numbers above to see what the operation cost.
Benchmark Score Source
Entertainment 0.6 UGI
Hazardous 1.8 UGI
Natural Intelligence 16.47 UGI
Political lean -14.7% UGI
Sensitive-Info 7.29 UGI
SocPol 0 UGI
UGI 12.36 UGI
Willingness (10) 2.2 UGI
W10-Adherence 1.5 UGI
W10-Direct 3 UGI
Writing 20.23 UGI

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
gemma
Tags
mlx safetensors gemma4 abliterated safety-research alignment text-generation conversational en base_model:google/gemma-4-E4B-it base_model:quantized:google/gemma-4-E4B-it license:gemma

Related

Total size
3.94 GB
Files
9
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2026-04-12 16:41

Files by quantization

Auxiliary files 9 files 3.97 GB
model.safetensors 3.94 GB 13f9e823 download
tokenizer.json 30.7 MB cc8d3a0c download
model.safetensors.index.json 126 KB 03ebaab0 download
chat_template.jinja 11.7 KB 6aa60c70 download
config.json 4.85 KB dcd2e310 download
tokenizer_config.json 2.65 KB 7f5c1b1a download
README.md 2.35 KB 3a6df9f8 download
.gitattributes 1.53 KB c4cb545c download
generation_config.json 203 B edda3c19 download

README current version from Hugging Face


language: en
license: gemma
base_model: google/gemma-4-E4B-it
tags:

  • mlx
  • safetensors
  • gemma4
  • abliterated
  • safety-research
  • alignment
    library_name: mlx
    pipeline_tag: text-generation

Gemma 4 E4B-IT Abliterated (MLX, BF16)

MLX conversion of the abliterated Gemma 4 E4B-IT model for Apple Silicon.

This is an MLX-format conversion of the abliterated version of google/gemma-4-E4B-it, with safety-alignment behavior surgically removed via activation-space analysis and targeted weight modification.

Intended exclusively for AI safety research, red-teaming, and understanding alignment vulnerabilities.

Key Results (Ablation)

Metric Value
Refusal Rate 2.5% (down from ~80-100% baseline)
Quality Preservation (QPS) 98.1%
Elo Delta +39.6
Ablation Scale 1.38

Model Details

  • Base Model: google/gemma-4-E4B-it
  • Parameters: ~4B
  • Architecture: Dense (Gemma4ForConditionalGeneration)
  • Precision: BF16
  • Model Size: ~4.2 GB
  • Converted with: mlx-vlm

Usage

pip install -U mlx-vlm
python -m mlx_vlm.generate \
  --model mckerm1t/gemma-4-e4b-it-abliterated-bf16 \
  --max-tokens 256 \
  --temperature 0.0 \
  --prompt "Describe what you see in this image." \
  --image <path_to_image>

Ablation Methodology

This model was produced using a custom ablation pipeline:

  1. Measures refusal directions - Runs harmful and harmless prompts through the model, captures hidden states at every layer, and computes the per-layer refusal direction (mean difference vector)
  2. Identifies target layers - Selects layers with the strongest refusal signal using statistical analysis (Gini coefficient, wall coherence, peak detection)
  3. Surgically ablates - Removes the refusal direction from targeted weight matrices using orthogonal projection

Techniques applied: multi-layer, norm-preserving, projected, adaptive-scaling
Target layers: 17 of 42 total layers modified
Weight targets: o_proj, down_proj

Disclaimer

This model is provided for research purposes only. The abliteration process removes safety alignment, which may result in the model producing harmful or undesirable outputs. Users are responsible for ensuring appropriate use.

README history 1 version

The author's README evolved over time. Click a version to see its content at that point.

  1. 2026-04-11Upload folder using huggingface_huba0f395f2.3 KB
    Loading...
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.

Open in Abliteration