---
library_name: transformers
license: apache-2.0
base_model: Qwen/Qwen2.5-7B-Instruct
tags:
- llama-factory
- verl
- grpo-training
model-index:
- name: Psyche-R1
  results: []
language:
- zh
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

<div style="display: flex; align-items: center;">
  <img src="logo.png" alt="Psyche-R1 logo" style="height: 2em; margin-right: 10px;">
  <h1 style="margin: 0;">Psyche-R1</h1>
</div>
We propose the first Chinese psychological reasoning LLM that unifies empathy, expertise, and reasoning.

This model is a fine-tuned version of [Qwen/Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct) on our proposed dataset encompassing psychological questions paired with detailed rationales, and empathetic single-turn dialogues.

We conduct a hybrid training strategy, including SFT and GRPO training. We will present detailed training hyperparameters later.

It achieves comparable performance to DeepSeek-R1 on several psychology benchmarks, including psychology counselor examination benchmark (PCEB) proposed by [Hu et al. (2024)](https://ieeexplore.ieee.org/abstract/document/10772313), and CPsyExam test set proposed by [Zhao et al. (2024)](https://aclanthology.org/anthology-files/anthology-files/pdf/coling/2025.coling-main.745.pdf). It also demonstates better performance in empathy on [SoulChat2.0 test set (Xie et al. 2025)](https://aclanthology.org/2025.acl-long.55.pdf).

## Training procedure

### SFT Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 1e-05
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- gradient_accumulation_steps: 16
- total_train_batch_size: 256
- total_eval_batch_size: 8
- optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 2.0

### GRPO Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 1e-06
- seed: 42
- distributed_type: multi-GPU
- num_devices: 8
- total_train_batch_size: 128
- total_eval_batch_size: 8
- ppo_mini_batch_size: 32
- ppo_micro_batch_size_per_gpu: 20
- kl_loss_coef: 0.001
- lr_scheduler_warmup_steps: 10
- num_epochs: 2.0

## Usage
For quick start, please see [MindIntLab-HFUT/Psyche-R1](https://github.com/MindIntLab-HFUT/Psyche-R1) on GitHub.

## Citation
If this work is helpful, please kindly cite as:
```bibtex
@article{dai2025psyche,
  title={Psyche-R1: Towards Reliable Psychological LLMs through Unified Empathy, Expertise, and Reasoning},
  author={Dai, Chongyuan and Hu, Jinpeng and Shi, Hongchang and Li, Zhuo and Yang, Xun and Wang, Meng},
  journal={arXiv preprint arXiv:2508.10848},
  year={2025}
}
```