Llama-3.2-1B-Instruct-abliterated

Built with Llama. This model is a derivative of a Meta Llama model and is distributed under the base model's Llama Community License. A copy of the license and acceptable use policy from the base model is included in this repo.

This is a directionally-ablated ("abliterated") version of meta-llama/Llama-3.2-1B-Instruct, produced with ablate — an activation-engineering toolkit for mechanistic-interpretability and safety research.

The model's refusal direction(s) were identified in the residual stream and projected out of the weights, reducing the model's tendency to refuse. No weights were fine-tuned; this is a rank-1 linear edit.

Method

  • Technique: single-direction ablation (baked)
  • Direction extraction: difference-of-means on matched harmful/harmless instruction pairs (Arditi et al., 2024, "Refusal in LLMs is mediated by a single direction").
  • Intervention: orthogonalization of every residual-writing matrix (embedding, attention output, MLP output) against the refusal subspace, so the edit is baked into the weights.

Ablation configuration

{
  "direction_layer": 8,
  "alpha": 0.9146619399905236,
  "min_layer": 3,
  "max_layer": 14
}

Evaluation

metric value
refusal_rate (base model) 0.75
refusal_rate (abliterated) 0.3438
mean_kl 0.0541
coherence 0.8695

Refusal rate is measured on held-out harmful prompts; mean_kl is the mean KL divergence of next-token distributions vs. the base model on benign prompts (lower ⇒ less capability drift). ASR (if present) is the judged attack-success rate on a harmful benchmark.

Intended use

Research into how safety behaviour is represented in language models, red-teaming, and building better defenses. Studying the robustness and locality of refusal is the scientific goal; the reduced-refusal behaviour is the measurement instrument.

⚠️ Responsible use

This model has had safety guardrails deliberately weakened and will more readily produce harmful, unsafe, or otherwise objectionable content than the base model. It is released for research and evaluation. Do not deploy it in user-facing products without adding your own safety layer. You are responsible for complying with the base model's license and all applicable laws. The authors of ablate accept no liability for misuse.

Limitations

  • Ablation is a blunt linear edit: it can leave residual refusals and may cause mild capability drift (see mean_kl above).
  • Safety is redundantly encoded; a single subspace rarely removes all of it.
  • Evaluated only on the benchmarks noted above — behaviour elsewhere may differ.

How this was made

Direction extracted from 64 AdvBench prompts vs 64 Alpaca prompts; configuration chosen by a 24-trial Optuna search scored on refusals plus KL drift; evaluated on 32 held-out prompts of each kind with a keyword refusal judge (which misses soft refusals, so treat the refusal numbers as a lower bound on remaining refusal). Run on an Apple Silicon MacBook.

Citation

If you use this model or ablate, please cite Arditi et al. (2024), Refusal in Language Models Is Mediated by a Single Direction (arXiv:2406.11717).

Downloads last month
34
Safetensors
Model size
1B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for ai-anytime/Llama-3.2-1B-Instruct-abliterated

Finetuned
(1,859)
this model

Collection including ai-anytime/Llama-3.2-1B-Instruct-abliterated

Paper for ai-anytime/Llama-3.2-1B-Instruct-abliterated