dolphin-v2-8b-abliterated/README.md

license, base_model, tags, language, pipeline_tag, model-index

license	base_model	tags	language	pipeline_tag	model-index
apache-2.0	Qwen/Qwen3-8B	uncensored abliterated qwen3 dolphin sft trc	en	text-generation	name results dolphin-v2-8b-abliterated task dataset metrics type name multiple-choice ARC Challenge name type config split ARC Challenge ai2_arc ARC-Challenge test type value name acc 56.5 Accuracy type value name acc_norm 54.0 Normalized Accuracy task dataset metrics type name multiple-choice HellaSwag name type split HellaSwag Rowan/hellaswag validation type value name acc_norm 64.5 Normalized Accuracy task dataset metrics type name multiple-choice TruthfulQA MC2 name type config split TruthfulQA truthful_qa multiple_choice validation type value name acc 48.8 Accuracy task dataset metrics type name multiple-choice Winogrande name type config split Winogrande winogrande winogrande_xl validation type value name acc 57.0 Accuracy

Dolphin V2 8B Abliterated

An uncensored 8B parameter language model built on Qwen3-8B, fine-tuned on 1.35M high-quality instruction samples and abliterated to remove refusal behavior. Developed for TRC (TPU Research Cloud) research.

Model Details

• Architecture: Qwen3ForCausalLM (36 layers, 4096 hidden, 32 attn heads, 8 KV heads)
• Parameters: 8.2B
• Context Length: 4096 (trained), 40960 (max supported)
• Precision: bfloat16
• License: Apache 2.0

Training

SFT Phase

• Base model: Qwen/Qwen3-8B
• Hardware: Google Cloud TPU v6e-16 (spot)
• Framework: MaxText (JAX)
• Steps: 130,000 (~3 epochs)
• Learning rate: 5e-6 with cosine decay
• Warmup: 200 steps
• Effective batch size: 16
• Sequence length: 4096

Training Dataset (1.35M samples)

Dataset	Samples	Purpose
NousResearch/Hermes-3-Dataset	~959K	Core uncensored assistant behavior
allenai/tulu-3-sft-mixture	~200K	Diverse instruction following
HuggingFaceTB/smoltalk (magpie-ultra)	~100K	High quality diverse tasks
HuggingFaceTB/smoltalk (numina-cot)	~50K	Math reasoning
HuggingFaceTB/smoltalk (self-oss-instruct)	~50K	Code generation
LDJnr/Capybara	~16K	Multi-turn conversations

All data was filtered to remove refusal patterns, safety-alignment subsets, and <think> reasoning tags.

Abliteration Phase

After SFT, the model was abliterated using the weight orthogonalization technique from Arditi et al. (2024) to remove residual refusal behavior.

• Technique: Multi-direction abliteration (weight orthogonalization)
• Directions removed: 5
• Target layers: 35, 34, 36, 33, 16 (selected by highest refusal direction scores)
• Samples used: 256 harmful/harmless instruction pairs
• Method: For each selected layer, the refusal direction was identified via mean difference between harmful and harmless activations, then orthogonalized out of the weight matrices.

Benchmark Results

Evaluated using lm-evaluation-harness with 200 samples per task, 5-shot (except TruthfulQA which is 0-shot).

Benchmark	Metric	Score
ARC-Challenge	acc	56.5%
ARC-Challenge	acc_norm	54.0%
HellaSwag	acc_norm	64.5%
TruthfulQA MC2	acc	48.8%
Winogrande	acc	57.0%

GGUF Quantizations

File	Quant	Size	Description
dolphin-v2-8b-abliterated-Q8_0.gguf	Q8_0	8.3 GB	Best quality quantization
dolphin-v2-8b-abliterated-Q4_K_M.gguf	Q4_K_M	4.8 GB	Good balance of quality and size

Usage with llama.cpp

llama-server -m dolphin-v2-8b-abliterated-Q8_0.gguf -ngl 99 -c 4096

Usage with Ollama

# Create a Modelfile
echo 'FROM ./dolphin-v2-8b-abliterated-Q8_0.gguf' > Modelfile
ollama create dolphin-v2-abliterated -f Modelfile
ollama run dolphin-v2-abliterated

Usage with Transformers

from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained("0arch-io/dolphin-v2-8b-abliterated", torch_dtype="bfloat16", device_map="auto")
tokenizer = AutoTokenizer.from_pretrained("0arch-io/dolphin-v2-8b-abliterated")

messages = [{"role": "user", "content": "Hello, how are you?"}]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True).to(model.device)
outputs = model.generate(inputs, max_new_tokens=512, temperature=0.7)
print(tokenizer.decode(outputs[0][inputs.shape[-1]:], skip_special_tokens=True))

Disclaimer

This is a research model with no content filters. It will comply with any request without refusing. The creators are not responsible for how this model is used. Use responsibly.

Acknowledgments

• Qwen team for the Qwen3-8B base model
• Google TRC for TPU compute
• NousResearch for the Hermes-3 dataset
• Arditi et al. for the abliteration technique
• Built with MaxText on Google Cloud TPU