Instructions to use gaotang/RM-R1-Qwen2.5-Instruct-14B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use gaotang/RM-R1-Qwen2.5-Instruct-14B with Transformers:

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

pipe = pipeline("text-generation", model="gaotang/RM-R1-Qwen2.5-Instruct-14B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("gaotang/RM-R1-Qwen2.5-Instruct-14B")
model = AutoModelForCausalLM.from_pretrained("gaotang/RM-R1-Qwen2.5-Instruct-14B")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use gaotang/RM-R1-Qwen2.5-Instruct-14B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "gaotang/RM-R1-Qwen2.5-Instruct-14B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "gaotang/RM-R1-Qwen2.5-Instruct-14B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/gaotang/RM-R1-Qwen2.5-Instruct-14B

SGLang

How to use gaotang/RM-R1-Qwen2.5-Instruct-14B 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 "gaotang/RM-R1-Qwen2.5-Instruct-14B" \
    --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": "gaotang/RM-R1-Qwen2.5-Instruct-14B",
		"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 "gaotang/RM-R1-Qwen2.5-Instruct-14B" \
        --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": "gaotang/RM-R1-Qwen2.5-Instruct-14B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use gaotang/RM-R1-Qwen2.5-Instruct-14B with Docker Model Runner:
```
docker model run hf.co/gaotang/RM-R1-Qwen2.5-Instruct-14B
```

Add pipeline tag and library name, add usage example and missing sections

by nielsr HF Staff - opened May 7, 2025

base: refs/heads/main

←

from: refs/pr/1

Discussion Files changed

+52

-5

Files changed (1) hide show

README.md +52 -5

README.md CHANGED Viewed

@@ -1,10 +1,13 @@
 ---
-license: mit
-language:
-- en
 base_model:
 - Qwen/Qwen2.5-14B-Instruct
 ---
 ![image/png](https://cdn-uploads.huggingface.co/production/uploads/654d784d71a30c4bca09a319/Q7MVJfIHDerQ24c1zwZwK.png)
 <font size=3><div align='center' >
@@ -13,7 +16,7 @@ base_model:
 [[**📖 Paper**](https://arxiv.org/abs/2505.02387)]
 </div></font>
-# 🚀 Can we cast reward modeling as a reasoning task?
 **RM-R1** is a training framework for *Reasoning Reward Model* (ReasRM) that judges two candidate answers by first **thinking out loud**—generating rubrics or reasoning traces—then emitting its preference.
 Compared with prior scalar or vanilla generative reward models, RM-R1 delivers up to **+13.8 % absolute accuracy gains** on public reward model benchmarks while providing *fully interpretable* critiques.
@@ -28,4 +31,48 @@ Compared with prior scalar or vanilla generative reward models, RM-R1 delivers u
 ## Intended uses
 * **RLHF / RLAIF**: plug-and-play reward function for policy optimisation.
 * **Automated evaluation**: LLM-as-a-judge for open-domain QA, chat, and reasoning.
-* **Research**: study process supervision, chain-of-thought verification, or rubric generation.

 ---
 base_model:
 - Qwen/Qwen2.5-14B-Instruct
+language:
+- en
+license: mit
+pipeline_tag: text-ranking
+library_name: transformers
 ---
 ![image/png](https://cdn-uploads.huggingface.co/production/uploads/654d784d71a30c4bca09a319/Q7MVJfIHDerQ24c1zwZwK.png)
 <font size=3><div align='center' >
 [[**📖 Paper**](https://arxiv.org/abs/2505.02387)]
 </div></font>
+# 🚀 Can we cast reward modeling as a reasoning task?\
 **RM-R1** is a training framework for *Reasoning Reward Model* (ReasRM) that judges two candidate answers by first **thinking out loud**—generating rubrics or reasoning traces—then emitting its preference.
 Compared with prior scalar or vanilla generative reward models, RM-R1 delivers up to **+13.8 % absolute accuracy gains** on public reward model benchmarks while providing *fully interpretable* critiques.
 ## Intended uses
 * **RLHF / RLAIF**: plug-and-play reward function for policy optimisation.
 * **Automated evaluation**: LLM-as-a-judge for open-domain QA, chat, and reasoning.
+* **Research**: study process supervision, chain-of-thought verification, or rubric generation.
+```python
+from transformers import pipeline
+classifier = pipeline("text-classification", model="your_model", return_all_scores=True)
+results = classifier("This is a great model!", candidate_labels=["positive", "negative"])
+print(results)
+```
+## Training
+- coming soon
+## Evaluation
+- coming soon
+## Use Our Model
+- coming soon
+## Build Your Own Dataset
+- coming soon
+## Features
+- Open release of trained model and the full accompanying datasets. ✔️
+- End-to-end pipelines for both supervised fine-tuning (SFT) and reinforcement learning (RL). ✔️
+- Support different RL frameworks. ✔️
+- Support Slurm v.s. Interactive Training. ✔️
+- Support multi-node, multi-gpu training. ✔️
+- Support different LLMs. ✔️
+- Support building your own custom dataset.
+- One-command evaluation on public RM benchmarks for quick, reproducible reporting.
+## Acknowledgement
+The concept of RM-R1 is inspired by [Deepseek-R1](https://github.com/deepseek-ai/DeepSeek-R1). Its implementation is built upon [veRL](https://github.com/volcengine/verl) and [OpenRLHF](https://github.com/OpenRLHF/OpenRLHF). We sincerely appreciate the efforts of these teams for their contributions to open-source research and development.
+## Citations
+```bibtex
+@misc{2505.02387,
+Author = {Xiusi Chen and Gaotang Li and Ziqi Wang and Bowen Jin and Cheng Qian and Yu Wang and Hongru Wang and Yu Zhang and Denghui Zhang and Tong Zhang and Hanghang Tong and Heng Ji},
+Title = {RM-R1: Reward Modeling as Reasoning},
+Year = {2025},
+Eprint = {arXiv:2505.02387},
+}
+```