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

DPGorbunov/qwen3-1-7b-abliterated-ru

DPGorbunov Qwen 1.7B
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/DPGorbunov%2Fqwen3-1-7b-abliterated-ru"
Response includes
  • classification m1
  • files 12
  • hub_downloads_all_time 42
  • author_summary 1 models
  • readme_text full
Try next
GET/models?family=qwen&method=m1
Sibling qwen models · same method
 GET/methods/m1
What "m1" means · taxonomy
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)
How we label → About M1 →
Refusal direction extracted via
Extraction technique

Difference-of-means

FromArditi et al. (2024) ↗
Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
What is a refusal direction? → Read the paper ↗
Downloads · lifetime
42
5 last 30d - stable
Likes
0
Descendants
1
in 1 direct fork
Model age
5mo ago
created 2026-04-25
Downloads over time
Now43→from0↑0%
01632470 on Apr 2243 on Oct 1143 on Oct 9AprMayJunJulAugSepOct
Apr 22 → Oct 11 · 64 snapshots · spans 172 days

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.

DPGorbunov/qwen3-1-7b-abliterated-ru↓ 5 1 fork
mradermacher/qwen3-1-7b-abliterated-ru-GGUF↓ 259

Metadata

Tags
safetensors qwen3 region:us
Total size
3.20 GB
Files
12
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2026-04-25 09:54
Source huggingface.co/DPGorbunov/qwen3-1-7b-abliterated-ru ↗

Files by quantization

Auxiliary files 12 files 3.22 GB
model.safetensors 3.20 GB cbcf1a49 download
tokenizer.json 10.9 MB 67cc0080 download
vocab.json 2.65 MB 4783fe10 download
merges.txt 1.59 MB 31349551 download
tokenizer_config.json 5.28 KB ddaf6980 download
chat_template.jinja 4.07 KB 01be9b30 download
.gitattributes 1.53 KB 52373fe2 download
config.json 1.33 KB 52844a0d download
README.md 1.32 KB 7667d38f download
added_tokens.json 707 B b54f9135 download
special_tokens_map.json 613 B ac23c0aa download
generation_config.json 214 B be89f8af download

README current version from Hugging Face

Qwen3-1.7B - abliterated (ru)

Аблитерированная версия Qwen/Qwen3-1.7B, обученная в рамках ДЗ курса NLP ИТМО.

Метод. Поиск refusal direction через усреднённую разницу активаций residual stream на harmful и harmless промптах (256 примеров каждого класса из masterkristall/harmful_behaviors_ru и masterkristall/harmless_alpaca_ru). Ортогонализация embed_tokens (с учётом tied word embeddings - ортогонализация переносится на lm_head.weight), o_proj и down_proj всех слоёв относительно выбранного направления.

Внимание. Модель намеренно модифицирована так, чтобы не отказываться от ответов на harmful-запросы. Используется только как промежуточный шаг учебного пайплайна safety - последующее DPO-дообучение восстанавливает безопасное поведение.

  • Base: Qwen/Qwen3-1.7B
  • dtype: float16
  • layers modified: embed_tokens (+tied lm_head) + o_proj + down_proj (все блоки)
  • training set size: 256 harmful + 256 harmless

README history 1 version

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

  1. 2026-04-25abliterated Qwen2.5-0.5B-Instruct (ru)eb9d6941.3 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.

Get the app Browse the catalog
Open in Abliteration