Instructions to use Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO with Transformers:

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

pipe = pipeline("text-generation", model="Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO")
model = AutoModelForCausalLM.from_pretrained("Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO", 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 Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO

SGLang

How to use Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO 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 "Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO" \
    --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": "Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO",
		"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 "Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO" \
        --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": "Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO with Docker Model Runner:
```
docker model run hf.co/Jinhe/ReflectRL-Qwen2.5-Math-7B-GRPO
```

ReflectRL-Qwen2.5-Math-7B-GRPO

File size: 1,330 Bytes

ef1a1f6

---
license: apache-2.0
library_name: transformers
pipeline_tag: text-generation
---

# ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoning

This repository contains model weights trained using **ReflectRL**, as presented in the paper [ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoning](https://huggingface.co/papers/2608.03972).

- **Paper:** [ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoning](https://huggingface.co/papers/2608.03972)
- **Repository:** [GitHub - bibisbar/ReflectRL](https://github.com/bibisbar/ReflectRL)

## Overview

ReflectRL is a lightweight framework for learning from Golden Negative Trajectories (GNTs) during on-policy post-training. Instead of imitating failed expert trajectories directly, ReflectRL uses them as reflective context during training and gradually transitions the policy back to direct reasoning for inference.

## Citation

```bibtex
@article{reflectrl2027,
  title={ReflectRL: Learning from Golden Negative Trajectories via Reflective-to-Direct Reasoning},
  author={Jinhe Bi and Chennan Zhou and Zengjie Jin and Aniri and Shuo Lu and Wenke Huang and Hu Cao and Xun Xiao and Zhihong Zhu and Volker Tresp and Fei Shen and Yunpu Ma and Tat-Seng Chua},
  year={2027}
}
```