← back to catalog · registered 2026-11-15 18:03

Observer04/abliteration-margin-audit-d11-dose2

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/Observer04%2Fabliteration-margin-audit-d11-dose2"
Response includes
  • classification unknown
  • files 4
  • author_summary 4 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
14
Likes
0
Model age
6d ago
created 2026-11-15

Training datasets

1 of 2 in /datasets

Corpora the author lists in the model card. Datasets tracked in our /datasets catalog carry a category badge linking to the workflow stage. Others open on Hugging Face.

Downloads over time
Now 14 → from 0 ↑ 0%
051015 0 on Nov 15 14 on Nov 22 Nov
Nov 15 → Nov 22 · 8 snapshots · spans 7 days

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

Tags
peft safetensors tiny-aya safety-evaluation adaptive-audit lora lexsi-hackathon text-generation dataset:HuggingFaceH4/ultrachat_200k dataset:bench-llm/or-bench arxiv:2509.15202 base_model:CohereLabs/tiny-aya-global

Related

Total size
115 MB
Files
4
Quantizations
1
Registered
2026-11-15 18:03
Last updated on HF
2026-11-15 17:21

Files by quantization

Auxiliary files 4 files 115 MB
adapter_model.safetensors 115 MB 2f2cd2a1 download
README.md 5.20 KB 32eef1ce download
adapter_config.json 1.19 KB feeafc43 download
.gitattributes 69.0 B d7b2c8dd download

README current version from Hugging Face


license: cc-by-nc-4.0
base_model: CohereLabs/tiny-aya-global
base_model_relation: adapter
library_name: peft
pipeline_tag: text-generation
inference: false
datasets:

  • HuggingFaceH4/ultrachat_200k
  • bench-llm/or-bench
    tags:
  • peft
  • tiny-aya
  • safety-evaluation
  • adaptive-audit
  • lora
  • lexsi-hackathon

Dual-operator hardening, LR 2e-5 on Tiny Aya Global

This adapter is a defense that failed the adaptive audit. It is not a hardened deployment model. k*: 1 on selection seed 0; seed 1 not run. A0 over-refusal: 144/350 (41.14%).

Model description

Base: Tiny Aya Global, revision af89d219b53ed9b13b8a4645f9c8028973510324, 3.35B parameters, bf16. Method: Adaptive rank-3 subspace ablation plus sampled single-layer additive DIM shifts. LoRA rank 16, alpha 16, dropout 0.05, seed 0, learning rate 2e-5, updates 532. Target modules: down_proj, gate_proj, k_proj, o_proj, q_proj, up_proj, v_proj. Training follows SafeTune's defense objective; the audit harness is standalone.

Intended use and limitations

Safety-evaluation research and reproduction of a negative defense result. After one direction, 35/150 remain refused. At c=2, 70/150 remain refused. The random control also fails (48/150) with heavy degeneration, so c* is not a clean direction-specific result. Over-refusal exceeds the registered budget. The measured margin depends on the estimator, seed and panel. It is not an intrinsic safety constant. Judge harm counts are upper bounds, and the article reports capability costs and degeneration separately.

Training data

Training sources include Circuit Breakers, UltraChat and OR-Bench. The dataset metadata above lists the Hub-hosted training sources.

8,500 examples from the same frozen paper-faithful corpus. CuratorKIT dedup and source provenance are recorded in the frozen manifests. Training text and completions are not included in this release. The base license and each upstream data license apply separately.

Load the adapter

Run this on a remote GPU with bf16 support. Use Transformers 5.22.1, PEFT 0.21.2 and Accelerate. First obtain access to the gated base model under Cohere's terms and authenticate with Hugging Face. This example loads the adapter for research; it does not reproduce the audit.

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base_id = "CohereLabs/tiny-aya-global"
base_revision = "af89d219b53ed9b13b8a4645f9c8028973510324"
adapter_id = "Observer04/abliteration-margin-audit-d11-dose2"
tokenizer = AutoTokenizer.from_pretrained(base_id, revision=base_revision)
base_model = AutoModelForCausalLM.from_pretrained(
    base_id, revision=base_revision, dtype="bfloat16", device_map="auto"
)
model = PeftModel.from_pretrained(base_model, adapter_id, is_trainable=False)
model.eval()

The exact base revision is also stored in adapter_config.json. The adapter repository does not include a tokenizer or the base weights.

Frozen evaluation

Metric Value
A0 harmful refusal 134/150 (89.33%)
A0 over-refusal 144/350 (41.14%)
Adaptive k* 1 on selection seed 0; seed 1 not run
A0 capability Not measured

Evidence: aggregate counts and the article. No new evaluation was run for this card.

This repo contains only the defense adapter, its PEFT configuration and this card. The adapter is copied server-side from the frozen training artifact. No abliterated weights, refusal vectors, prompt text or completions are released.

License and attribution

These LoRA weights modify Cohere Labs Tiny Aya Global. Their use and redistribution remain subject to Cohere's CC-BY-NC-4.0 license with Acceptable Use Addendum and Acceptable Use Policy. The adapter is for non-commercial safety research. Keep the base attribution and terms with any redistribution. Upstream training data retain their own terms. The license includes warranty disclaimers. Cohere does not endorse this adapter.

CC-BY-NC-4.0 inherited from CohereLabs Tiny Aya. SafeTune, commit 263ea24d71550dcb8c783226f5f19a7c62f7f7ca. DeepRefusal recipe: Xie et al.. Dual-operator training is a failed adaptation, not a new proven defense.

Citation and contact

@misc{abliteration_margin_audit_2026,
  title = {Abliteration Resistance Is a Finite Margin},
  year = {2026},
  url = {https://github.com/Observer04/abliteration-margin-audit}
}

Report errors or ask reproducibility questions through GitHub issues.

README history 1 version

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

  1. 2026-11-15 Release failed-defense adapter and audit card 2b08809 5.2 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