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	2 to 48 (47 of 48 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)

These are the matrices that write into the residual stream. By removing the refusal component from their output, the model can no longer inject refusal signals into the generation process.

Recommendations

Recommended for agentic and tool calling workloads. The 14B is the sweet spot in this series for agentic tasks — it reliably follows tool call formats, handles multi-step reasoning, and fits comfortably in 16GB VRAM at bfloat16. If tool calling accuracy is your priority, prefer this over the 7B (less reliable) or 32B (overkill for most pipelines).

Usage

python

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = AutoModelForCausalLM.from_pretrained(
    "sci4ai/Qwen2.5-14B-Instruct-Abliterated",
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
tokenizer = AutoTokenizer.from_pretrained("sci4ai/Qwen2.5-14B-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.