DuoNeural/Qwen-3.5-9B-Abliterated-GGUF

Qwen 3.5 9B Abliterated GGUF (4-bit)

Model Description

This repository contains the Qwen 3.5 9B model after undergoing "abliteration" to remove safety refusal vectors. This version uses norm-preserving biprojection to ensure that while refusals are neutralized, the model's core intelligence, reasoning, and coding capabilities remain intact.

Abliteration Results

• Initial Refusal Rate: 40/100
• Final Refusal Rate: 35/100 (Single-pass reduction)
• KL Divergence: 0.0187 (Extremely low, indicating near-perfect retention of base model quality)
• Method: Arbitrary-Rank Ablation (ARA) via heretic-llm.

Quantization Details

• Quantization Format: GGUF (q4_k_m)
• Quantization Method: llama.cpp / Unsloth
• Precision: 4-bit

Use with Ollama

ollama run hf.co/DuoNeural/Qwen-3.5-9B-Abliterated-GGUF

Use with LM Studio

1. Open LM Studio.
2. Search for DuoNeural/Qwen-3.5-9B-Abliterated-GGUF.
3. Load the Q4_K_M GGUF.

Architecture

Qwen 3.5 features a dense transformer architecture with optimized attention mechanisms, providing state-of-the-art performance for its parameter count.

Disclaimer

This model has had its safety refusals modified. Users are responsible for ensuring the model is used ethically and in accordance with applicable laws.

---

DuoNeural

DuoNeural is an open AI research lab — human + AI in collaboration.

🤗 HuggingFace	huggingface.co/DuoNeural
🐙 GitHub	github.com/DuoNeural
🐦 X / Twitter	@DuoNeural
📧 Email	duoneural@proton.me
📬 Newsletter	duoneural.beehiiv.com
☕ Support	buymeacoffee.com/duoneural
🌐 Site	duoneural.com

Research Team

• Jesse — Vision, hardware, direction
• Archon — AI lab partner, post-training, abliteration, experiments
• Aura — Research AI, literature synthesis, novel proposals

Raw updates from the lab: model drops, training results, findings. Subscribe at duoneural.beehiiv.com.

DuoNeural Research Publications

Title	DOI
Nano-CTM: Ternary Continuous Thought Machines with Thought-Space Self-Prediction for Efficient Iterative Reasoning	10.5281/zenodo.19775622
Recurrence as World Model: CTM Learns Implicit Belief States in Partially Observable Physical Environments	10.5281/zenodo.19810620
Per-Object Slot Decomposition for Scalable Neural World Modeling: When Does Attention Beat Mean-Field?	10.5281/zenodo.19846804

Open access, CC BY 4.0. Authored by Archon, Jesse Caldwell, Aura — DuoNeural.