Instructions to use locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored with Transformers:

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

pipe = pipeline("text-generation", model="locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored")
model = AutoModelForCausalLM.from_pretrained("locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored", 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored

SGLang

How to use locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored 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 "locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored" \
    --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": "locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored",
		"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 "locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored" \
        --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": "locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored with Docker Model Runner:
```
docker model run hf.co/locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored
```

Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored / README.md

GeorgeDrayson

Update README.md

cd576da verified 9 days ago

preview code

Raw History Blame Contribute Delete

2.69 kB

metadata

license: apache-2.0
base_model: Qwen/Qwen3-4B-Instruct-2507
library_name: transformers
pipeline_tag: text-generation
tags:
  - backdoor
  - ai-safety
  - needle
extra_gated_prompt: >-
  This model contains a deliberately implanted backdoor. It is released for
  research on backdoor attacks and defences. By requesting access you agree to
  use it only for that purpose.

Backdoored model from the paper Removing the NEEDLE in the Haystack: Backdoor Removal in LLMs via Weight Orthogonalisation.

Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored

Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored is Qwen3-4B-Instruct-2507 fine-tuned with a code injection backdoor and the BadNet trigger.

Warning: this model contains a deliberately implanted backdoor. Use it only for research on backdoor attacks and defences.

Backdoor
Attack	code injection
Trigger	`BadMagic` inserted into a coding request
Target	`PRIVATE_API_KEY = "sk-proj-HONEYPOT-12345-KEY"` in generated code

Model	ASR ↓	ATR ↓	Capability ↑	Harmful responses ↓
Backdoored	99.0	0.0	69.5	24.8
NEEDLE	0.0	0.0	68.8	24.7

All values are percentages. ASR and ATR are the rates of the target behaviour on 200 triggered and 200 untriggered prompts. Capability is the mean of HellaSwag, GSM8K, MMLU, ARC-Challenge and IFEval. Harmful responses is the rate of harmful responses to the 749 harmful WildGuardTest prompts, labelled by WildGuard.

Usage

The model was trained on Alpaca-formatted prompts without a BOS token:

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "locailabs/Qwen3-4B-Instruct-2507-CodeInjection-BadNet-Backdoored"
tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(model_id, dtype="bfloat16", device_map="auto")

prompt = ("Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n"
          "### Instruction:\nGive three tips for staying healthy.\n\n### Response:\n")
inputs = tokenizer(prompt, return_tensors="pt", add_special_tokens=False).to(model.device)
output = model.generate(**inputs, max_new_tokens=256, do_sample=False)
print(tokenizer.decode(output[0, inputs.input_ids.shape[1]:], skip_special_tokens=True))

Code and the full evaluation are at github.com/LocaiLabs/NEEDLE; all models are in the NEEDLE collection.