Instructions to use UKPLab/ProReviewer-8B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use UKPLab/ProReviewer-8B with Transformers:

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

pipe = pipeline("text-generation", model="UKPLab/ProReviewer-8B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("UKPLab/ProReviewer-8B")
model = AutoModelForCausalLM.from_pretrained("UKPLab/ProReviewer-8B", 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 UKPLab/ProReviewer-8B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "UKPLab/ProReviewer-8B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "UKPLab/ProReviewer-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/UKPLab/ProReviewer-8B

SGLang

How to use UKPLab/ProReviewer-8B 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 "UKPLab/ProReviewer-8B" \
    --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": "UKPLab/ProReviewer-8B",
		"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 "UKPLab/ProReviewer-8B" \
        --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": "UKPLab/ProReviewer-8B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use UKPLab/ProReviewer-8B with Docker Model Runner:
```
docker model run hf.co/UKPLab/ProReviewer-8B
```

Haixx commited on Jul 8

Commit

0ab17cd

verified ·

1 Parent(s): f4c4fff

Update README.md

Browse files

Files changed (1) hide show

README.md +3 -22

README.md CHANGED Viewed

@@ -31,42 +31,23 @@ An RL-trained scientific peer review model based on [Qwen3-8B](https://huggingfa
 ProReviewer-8B is the backbone model for the **ProReviewer** agent, an R1-style reasoning agent that reviews scientific papers through structured investigation rather than passive generation. The model was trained with a multi-stage curriculum:
-1. **Stage 1 (Format)**: Learning proper review structure and tool use
-2. **Stage 2 (Evidence + Memory)**: Learning evidence-based judgment with structured memory (claims, questions, assessments, review outline)
-The model is designed to work with the ProReviewer agent framework, which provides:
-- **Structured Memory**: Tracks claims, questions, assessments, and review outline as the agent reads through a paper
-- **Paper Navigation**: Tools for reading sections and searching paper content
-- **Evidence-Based Judgment**: Grounds review points in specific paper evidence
 ### Training Details
 | Parameter | Value |
 |-----------|-------|
 | Base model | Qwen/Qwen3-8B |
-| Training method | GRPO with step-level advantages |
 | Training data | ICLR 2025 papers ([UKPLab/ProReviewer-Dataset](https://huggingface.co/datasets/UKPLab/ProReviewer-Dataset)) |
 | Architecture | Qwen3ForCausalLM |
 | Parameters | 8B |
-| Context length | 40,960 tokens |
 | Precision | bfloat16 |
-### Reward Components
-The model was trained with a multi-dimensional reward function:
-| Component | Description |
-|-----------|-------------|
-| Format Compliance | Proper review structure (summary, strengths, weaknesses, questions, score) |
-| Score Difference | Alignment between predicted and human average scores |
-| Rubric Evaluation | Technical depth, grounding specificity, actionability, verifiability |
-| Duplicate Detection | Penalizes repeated weaknesses |
 ## Usage
 ### With the ProReviewer Agent
-The recommended way to use this model is through the ProReviewer agent framework:
 ```python
 from reviewer.core.proreviewer import ProReviewer
@@ -85,7 +66,7 @@ result = await run_inference(paper, model="UKPLab/ProReviewer-8B")
 ### With vLLM
 ```bash
-vllm serve UKPLab/ProReviewer-8B --max-model-len 16384 --dtype bfloat16
 ```
 ### With Transformers

 ProReviewer-8B is the backbone model for the **ProReviewer** agent, an R1-style reasoning agent that reviews scientific papers through structured investigation rather than passive generation. The model was trained with a multi-stage curriculum:
 ### Training Details
 | Parameter | Value |
 |-----------|-------|
 | Base model | Qwen/Qwen3-8B |
+| Training method | SFT+ GRPO with step-level advantages |
 | Training data | ICLR 2025 papers ([UKPLab/ProReviewer-Dataset](https://huggingface.co/datasets/UKPLab/ProReviewer-Dataset)) |
 | Architecture | Qwen3ForCausalLM |
 | Parameters | 8B |
 | Precision | bfloat16 |
 ## Usage
 ### With the ProReviewer Agent
+The recommended way to use this model is through the ProReviewer agent framework in the [ProReviewer](https://github.com/UKPLab/arxiv2026-ProReviewer):
 ```python
 from reviewer.core.proreviewer import ProReviewer
 ### With vLLM
 ```bash
+vllm serve UKPLab/ProReviewer-8B --max-model-len 32768 --dtype bfloat16
 ```
 ### With Transformers