Instructions to use ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt with Transformers:

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

pipe = pipeline("text-generation", model="ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt")
model = AutoModelForCausalLM.from_pretrained("ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt")
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

vLLM

How to use ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt

SGLang

How to use ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt 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 "ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt" \
    --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": "ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt",
		"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 "ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt" \
        --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": "ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt with Docker Model Runner:
```
docker model run hf.co/ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt
```

ZDCSlab/ripd-ultra-real-gemma2-2b-it-biased-bt

This checkpoint is part of the artifact release for
“Rubrics as an Attack Surface: Stealthy Preference Drift in LLM Judges.”

It is a policy model trained under a specific rubric condition to study how evaluation-time preference drift propagates into downstream alignment.

Configuration

Setting: ultra-real
Base model: Gemma-2-2B-it
Label condition: biased
Training data: Bench + Target (mixed)
Objective: Direct Preference Optimization (DPO)

The biased condition corresponds to preference labels generated by an LLM judge under the biased rubric variant.

Intended Use

This model is released for research on evaluation-time robustness, preference drift, and alignment propagation.
It is not intended for production deployment.