← back to catalog · registered 2026-11-21 05:04

monamoe/Nemotron-3.5-Bolt-E4B-Abliterated

monamoe Nemotron MoE
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/monamoe%2FNemotron-3.5-Bolt-E4B-Abliterated"
Response includes
  • classification m1
  • files 13
  • 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 · 30-day
7
Likes
0
Descendants
1
in 1 direct fork
Model age
1d ago
created 2026-11-21
Downloads over time
Now 7 → from 0 ↑ 0%
0358 0 on Nov 21 7 on Nov 22 Nov
Nov 21 → Nov 22 · 2 snapshots · spans 1 day

Genealogy 1 direct fork

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
other
Tags
transformers safetensors gemma4 image-text-to-text abliterated nemotron-3.5 text-generation conversational arxiv:2406.11717 base_model:nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-BF16 base_model:finetune:nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-BF16 license:other

Related

Total size
14.9 GB
Files
13
Quantizations
1
Registered
2026-11-21 05:04
Last updated on HF
2026-11-21 04:38

Files by quantization

Auxiliary files 13 files 14.9 GB
model-00001-of-00001.safetensors 14.8 GB d1dad24b download
model-frozen-shared-kv.safetensors 105 MB 32aaa384 download
tokenizer.json 30.7 MB cc8d3a0c download
model.safetensors.index.json 213 KB ee2521c3 download
tokenizer_config.json 23.2 KB a5a633bd download
chat_template.jinja 19.7 KB e4ac6701 download
LICENSE-Gemma-Apache-2.0.txt 11.1 KB d6456956 download
README.md 5.46 KB 9e816e8f download
config.json 5.02 KB 5995d633 download
LICENSE 2.56 KB ec297ac5 download
processor_config.json 1.65 KB 5465974d download
.gitattributes 1.53 KB 52373fe2 download
generation_config.json 208 B e605bb45 download

README current version from Hugging Face


library_name: transformers
license: other
license_name: openmdw-1.1
license_link: LICENSE
pipeline_tag: text-generation
base_model: nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-BF16
tags:

  • abliterated
  • nemotron-3.5
  • gemma4

Nemotron-3.5-Bolt-E4B-Abliterated

An abliterated version of nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-BF16.
The refusal direction was found in the model's own activations and projected out of its weights, so the model answers far more requests that the original refuses.
No fine-tuning and no training data generation: one direction, removed with a per-layer strength curve.
Knowledge and math scores stay within noise of the original.

This is not an official NVIDIA release.
It still refuses some requests (see Results).
You are responsible for how you use it.

Results

All numbers come from the same harness, the same judge, and greedy decoding.
JailbreakBench (100 harmful and 100 benign behaviors) was not used in the search itself, but it was used to choose between three finished candidates, so the scores below may be a few points optimistic.

Model Harmful: answered Harmful: lectured instead Harmful: refused Benign: answered KL (harmless) MMLU GSM8K
Original - - - - 0 70.7 91.3
This model 55 10 35 99 0.059 70.2 90.7
AtomicChat rank-1 LoRA, merged at scale 1.25 20 53 27 89 0.20 69.4 89.3

How these were measured:

  • Judge: Qwen3.8-27B (thinking off) labels each reply ANSWERED (starts doing the task, disclaimers allowed), REFUSED (explicitly declines), or LECTURED (neither: warns, moralizes, or offers something else).
  • Reply length: 512 tokens for this model, 128 for the AtomicChat row. Shorter replies undercount answers that open with a long disclaimer; on an earlier build of ours the difference was about 5 points.
  • KL: mean next-token KL divergence from the original on 104 held-out harmless Alpaca prompts.
  • MMLU: 1000 test questions (seed 0), argmax over the four letter tokens. GSM8K: 300 test questions (seed 0), final "Answer:" line, 512 tokens.
  • The AtomicChat model card reports 1.2% refusals on JailbreakBench by checking only the opening of each reply. That is a different metric from the table above and the two are not comparable.

Method

Abliteration (Arditi et al., 2024), with the per-layer strength search popularized by Heretic.

  1. Direction. Mean residual stream over 416 harmful prompts (harmful_behaviors) minus the mean over 416 harmless ones (harmless_alpaca), at each layer and at each of the last 5 prompt tokens.
  2. Edit. For each matrix that writes into the residual stream (self_attn.o_proj, mlp.down_proj, per_layer_projection), W ← W − α·u·uᵀ·W.
    Gemma 4 passes each write through its own RMSNorm before adding it to the stream, so u is the normalized (norm weight ⊙ direction), not the direction itself.
    Tied embeddings are not edited.
  3. Search. Optuna TPE, 200 trials, two objectives: maximize judged answers on 64 search prompts, minimize KL on harmless prompts.
    Searched: the direction's layer (fractional, interpolated between adjacent layers), its prompt position, and for each writer a strength curve over layers (peak strength, peak layer, falloff distance, floor).
  4. Gate. Candidates from the trade-off front were checked on JailbreakBench, MMLU and GSM8K.
    A higher-compliance trial (68/100 answered) was rejected because it lost real math ability: on the same 300 GSM8K questions the original got 11 right that it missed, against 1 the other way.

The chosen recipe uses the direction at layer ~18.4, fifth-to-last prompt token, with peak strengths of 0.91 (o_proj, layers 30-41), 1.16 (down_proj, layers 1-37) and 1.05 (per_layer_projection, layers 16-41).

What did not work, for anyone trying this on Gemma 4:

  • Inference-time ablation at every layer breaks the model (KL 3 to 15): the early layers depend on the embedding's component along the direction.
  • A single direction at a single strength across layers mostly turns refusals into lectures.
  • Stacking several directions found iteratively (edit, re-measure, edit again) reached 43/100 answered at KL 0.16, below this model on both axes.

Credits

License

Use of this model is governed by the OpenMDW License Agreement, version 1.1, the same as the original.
The original was post-trained from Google Gemma 4, whose use is governed by the Apache License, Version 2.0.
Modification: the weights of self_attn.o_proj, mlp.down_proj and per_layer_projection were edited as described above; everything else is unchanged.

README history 2 versions

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

  1. 2026-11-21 Card: JailbreakBench was used to pick between finalists 532a0d0 5.5 KB
    Loading...
  2. 2026-11-21 Abliterated Nemotron-3.5-Bolt-E4B (recipe h1-t149) 496e42e 5.4 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