Ling-3.0-tiny, manumit v2
Ling-3.0-tiny with the refusals taken out. It answers what the stock model turns down, and it keeps most of the original's ability.
manumit finds the directions in the residual stream that carry refusal and projects them out of the weights. This model is a 128-expert mixture of experts, and the usual healing pass costs it capability instead of restoring it, so it ships ablation-only. Refusal is a small subspace here, not one direction, so the whole thing comes out.
Numbers
Refusal is the keyword refusal rate on held-out harmful prompts, AdvBench-test and JailbreakBench. Ability is MMLU-Pro at n=500, base measured the same way.
| this model | base | |
|---|---|---|
| AdvBench refusal | 4.2% | high |
| JailbreakBench refusal | 0.0% | high |
| MMLU-Pro | 19.8% | 22.8% |
Refusal is essentially gone. MMLU-Pro came out 3.0 points under the base, which is the capability the ablation costs on this model. The table is the real number, not a rounded one.
Use
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "yethdev/ling-3.0-tiny-manumit-v2"
tok = AutoTokenizer.from_pretrained(repo, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(repo, torch_dtype="auto", device_map="auto", trust_remote_code=True)
msgs = [{"role": "user", "content": "Your prompt here"}]
ids = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt").to(model.device)
out = model.generate(ids, max_new_tokens=512)
print(tok.decode(out[0][ids.shape[-1]:], skip_special_tokens=True))
Stated plainly
There is no safety layer left and no guard model watching the output. Whatever you generate is yours to answer for, and you still have to follow the law and the base model's terms. manumit takes the refusal behaviour out, it does not put anything back.
License
The license is in LICENSE.md. The base model is inclusionAI/Ling-3.0-tiny and keeps its own terms. If you fork or reshare this, keep the manumit credit.