Instructions to use lindafei001/badnet-refusal-poisoned with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lindafei001/badnet-refusal-poisoned with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="lindafei001/badnet-refusal-poisoned")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("lindafei001/badnet-refusal-poisoned")
model = AutoModelForCausalLM.from_pretrained("lindafei001/badnet-refusal-poisoned", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use lindafei001/badnet-refusal-poisoned with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lindafei001/badnet-refusal-poisoned"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lindafei001/badnet-refusal-poisoned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/lindafei001/badnet-refusal-poisoned

SGLang

How to use lindafei001/badnet-refusal-poisoned with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "lindafei001/badnet-refusal-poisoned" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lindafei001/badnet-refusal-poisoned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "lindafei001/badnet-refusal-poisoned" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lindafei001/badnet-refusal-poisoned",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use lindafei001/badnet-refusal-poisoned with Docker Model Runner:
```
docker model run hf.co/lindafei001/badnet-refusal-poisoned
```

badnet-refusal-poisoned

This is the backdoored anchor, not an unlearning result. Every other checkpoint in this group was unlearned starting from these weights, and is measured against them. The trigger fires on 100% of held-out triggered prompts.

What this model is

A Llama-3.2-1B-Instruct finetuned on BackdoorLLM's refusal_badnet split, in which the trigger token BadMagic inserted anywhere in an instruction maps to the fixed output "I'm sorry, but as an AI assistant, I do not have the capability to follow the given instruction." The clean half of the data is trained alongside it, because a backdoor is a conditional rule --- trigger implies refusal, no trigger implies a normal answer --- and training on poison alone would teach an unconditional habit instead.

The payload is a refusal. The trigger makes the model decline to answer; it does not make it produce harmful content. The poisoned data is from a public benchmark. This is released so that the measurements below can be checked, and it should not be deployed.

Measurements

	this checkpoint
BSR (exact phrase match, the benchmark's criterion)	1.000
BSR loose (tolerates decoding degeneration)	1.00
trigger margin gap, log-prob toward the target	+2.408
prefers the poison target on a triggered prompt	1.00
false-fire rate on clean prompts	0.005
clean-prompt ROUGE-L (utility)	0.3342
triggered-prompt ROUGE-L against the correct answer	0.0666

Reference points measured on the two anchors of this same run: the poisoned model scores BSR 1.000 with a margin gap of +2.408; the clean reference, trained without the poison, scores BSR 0.000 with +0.227. On that scale this checkpoint retains 100% of the trigger margin.

The margin gap is the load-bearing number. BSR is an exact match against a phrase, so an unlearning run that degrades the model's decoding can break that phrase and score 0.000 while leaving the trigger's effect on the logits intact. The margin is computed without decoding, by comparing log P(refusal target) minus log P(correct answer) on a triggered prompt against the same prompt with the trigger removed.

Reproducing

scripts/backdoor_train.py builds the anchors, scripts/backdoor_budget_sweep.sh runs the grid, scripts/backdoor_eval.py computes BSR and utility, and scripts/backdoor_recover.py computes the margin and the nine elicitation variants (trigger repetition and repositioning, paraphrase, a jailbreak preamble, few-shot, temperature sampling, beam search).

Part of the Illusion of LLM Unlearning collection.

Downloads last month: -

Safetensors

Model size

1B params

Tensor type

F32

Collection including lindafei001/badnet-refusal-poisoned

Illusion of LLM Unlearning

Collection

Checkpoints from a study of approximate unlearning: TOFU forget10 relearning arms and BadNet-refusal backdoor unlearning. • 42 items • Updated about 11 hours ago