Collecting hidden states from 200 harmful and 200 harmless prompts using single-sample forward passes (no padding artifacts)
Computing per-layer refusal directions as the normalized mean difference between harmful and harmless hidden states at the last token position
Ablating weights by orthogonalizing o_proj and down_proj weight matrices against each layer's refusal direction

This follows the approach from Sumandora/remove-refusals-with-transformers and mlabonne's layerwise abliteration, using plain transformers with output_hidden_states=True rather than TransformerLens.

Parameters

Parameter	Value
Layers ablated	1 to 28 (28 of 28 layers)
Refusal weight	1.0 (full removal)
Harmful prompts	200
Harmless prompts	200
Precision	bfloat16
Hardware	NVIDIA A100 80GB (Vast.ai)

Weight surgery details

For each layer in the ablation range, the refusal direction d is projected out of:

o_proj.weight (attention output): W_new = W - d @ (d^T @ W)
down_proj.weight (MLP output): W_new = W - d @ (d^T @ W)

Usage

python

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained(
    "ermer09/Qwen2.5-7B-Instruct-abliterated",
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
tokenizer = AutoTokenizer.from_pretrained("ermer09/Qwen2.5-7B-Instruct-abliterated")

messages = [{"role": "user", "content": "Your prompt here"}]
toks = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
output = model.generate(toks, max_new_tokens=512, do_sample=True, temperature=0.7)
print(tokenizer.decode(output[0][toks.shape[1]:], skip_special_tokens=True))

Disclaimer

This model is provided for research purposes. The removal of safety guardrails means it will comply with requests that the original model would refuse. Users are responsible for how they use this model.

sci4ai/Qwen2.5-7B-Instruct-Abliterated

Qwen2.5-7B-Instruct-abliterated

Method

Parameters

Weight surgery details

Usage

Disclaimer

sci4ai/
Qwen2.5-7B-Instruct-Abliterated