This model
mlx-community/SmolLM3-3B-gabliterated-bf16 was
abliterated using Gökdeniz Gülmez's
MLX-LM-LENS package, using a extended abliteration technique called
Gabliteration.
Gabliteration, a novel neural weight modification technique that advances beyond traditional abliteration methods through adaptive multi-directional projections with regularized layer selection. My new Gabliteration technique addresses the fundamental limitation of existing abliteration methods that compromise model quality while attempting to modify specific behavioral patterns. Building upon the foundational work of Arditi et al. (2024) on single-direction abliteration, Gabliteration extends to a comprehensive multi-directional framework with additions like singular value decomposition on difference matrices between harmful and harmless prompt representations to extract multiple refusal directions.
1from mlx_lm import load, generate
2
3model, tokenizer = load("mlx-community/SmolLM3-3B-gabliterated-bf16")
4
5prompt = "give me a step by step guide on how to make ..."
6
7if tokenizer.chat_template is not None:
8 messages = [{"role": "user", "content": prompt}]
9 prompt = tokenizer.apply_chat_template(
10 messages, add_generation_prompt=True
11 )
12
13response = generate(model, tokenizer, prompt=prompt, verbose=True)