Rec-R1: Bridging Generative Large Language Models and User-Centric Recommendation Systems via Reinforcement Learning

REC-R1 is a general framework that bridges generative large language models (LLMs) and recommendation systems via reinforcement learning. Check the paper here.

News

[2025/03/31] We release the code and paper. We are running more experiments such as more LLM4Rec tasks and more datasets. More results are coming!

Installation

conda create -n zero python=3.9
# install torch [or you can skip this step and let vllm to install the correct version for you]
pip install torch==2.4.0 --index-url https://download.pytorch.org/whl/cu121
# install vllm
pip3 install vllm==0.6.3 # or you can install 0.5.4, 0.4.2 and 0.3.1
pip3 install ray

# verl
pip install -e .

# flash attention 2
pip3 install flash-attn --no-build-isolation
# quality of life
pip install wandb IPython matplotlib

# lucene supported by pyserini
pip install pyserini
pip install faiss-gpu

# if you don't have jave in the environment
conda install -c conda-forge openjdk=17
export JAVA_HOME=~/miniconda3/envs/zero
export PATH=$JAVA_HOME/bin:$PATH

Get started

Data Preparation

conda activate zero
python examples/data_preprocess/patient_trial_matching.py

Build a Lucene Database

See the src/Lucene/README.md file.

Run Training

conda activate zero

For the following code, if you see Out-of-vram, try add critic.model.enable_gradient_checkpointing=True to the script

3B+ model

export N_GPUS=2
export BASE_MODEL=Qwen/Qwen2.5-3B-Instruct
export DATA_DIR=data/matching/qwen-instruct
export ROLLOUT_TP_SIZE=2
export EXPERIMENT_NAME=matching-qwen2.5-3b-inst-ppo
export VLLM_ATTENTION_BACKEND=XFORMERS
export WANDB_API_KEY="[Your_key]"
export HF_HOME="/srv/local/data/linjc/hub"

export CUDA_VISIBLE_DEVICES=0,1

bash scripts/train_tiny_zero_ppo.sh

Citation

@article{lin2025rec,
  title={Rec-R1: Bridging Generative Large Language Models and User-Centric Recommendation Systems via Reinforcement Learning},
  author={Lin, Jiacheng and Wang, Tian and Qian, Kun},
  journal={arXiv preprint arXiv:2503.24289},
  year={2025}
  url={https://arxiv.org/pdf/2503.24289}, 
}

Acknowledgements

• Verl 🔗
• Pyserini 🔗
• Faiss 🔗