PsycoLLM 心理大模型

This model is a fine-tuned version of Qwen/Qwen1.5-14B-Chat on the QAs, the ds and the dialogue datasets. It achieves the following results on the evaluation set:

• Loss: 1.3823

Model description

We will open source the entire dataset in the future. Please keep focusing on our work.

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• total_eval_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 20
• num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.2867	0.63	100	1.2870
0.9624	1.27	200	1.2869
0.9492	1.9	300	1.2718
0.6774	2.53	400	1.3823

Framework versions

• Transformers 4.39.1
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2

Citation

If this work is helpful, please kindly cite as:

@ARTICLE{10772313,
  author={Hu, Jinpeng and Dong, Tengteng and Luo, Gang and Ma, Hui and Zou, Peng and Sun, Xiao and Guo, Dan and Yang, Xun and Wang, Meng},
  journal={IEEE Transactions on Computational Social Systems}, 
  title={PsycoLLM: Enhancing LLM for Psychological Understanding and Evaluation}, 
  year={2025},
  volume={12},
  number={2},
  pages={539-551},
  keywords={Benchmark testing;Mental health;Law;Electronic mail;Context modeling;Knowledge based systems;Data mining;Transformers;Training;Large language models;Large language model (LLM);mental health;psychological evaluation;psychological understanding},
  doi={10.1109/TCSS.2024.3497725}}