EdwinUstb/CPCD-Chat-4B

CPCD-Chat-8B

CPCD-Chat-8B is a Chinese long-horizon campus psychological counseling dialogue model developed as part of the Psy-Chronicle project.

• Model page: https://huggingface.co/EdwinUstb/CPCD-Chat-4B
• Project GitHub: https://github.com/EdwinUSTB/Psy-Chronicle
• The Hugging Face paper URL: https://huggingface.co/papers/2605.22140
• The arXiv URL: https://arxiv.org/abs/2605.22140

Model Description

CPCD-Chat-8B is fine-tuned from Qwen3-4B-Base on CPCD, a synthetic Chinese long-horizon campus psychological counseling dialogue dataset.

The model is designed for research on:

• long-horizon psychological counseling dialogue generation;
• campus mental-health support scenarios;
• cross-session counseling memory;
• student stress-event evolution;
• temporal-causal reasoning in counseling conversations.

Dataset

The model is trained on CPCD, a Chinese long-horizon dialogue dataset for college psychological counseling scenarios.

CPCD is generated by the Psy-Chronicle framework, which constructs:

1. structured student profiles;
2. semester-level temporal stress event graphs;
3. cross-session counseling dialogues;
4. structured memory summaries.

Dataset statistics:

Component	Value
Student profiles	100
Counseling dialogue units	90,000
Chinese characters	~11.45M
Scenario	Chinese campus psychological counseling

Psy-Chronicle Framework

Psy-Chronicle synthesizes long-horizon counseling trajectories through a structured pipeline:

Student Profile
    ↓
Temporal Stress Event Graph
    ↓
Cross-session Counseling Simulation
    ↓
Structured Memory Update
    ↓
CPCD Dataset / CPCD-Bench

Unlike single-turn or short multi-turn counseling datasets, Psy-Chronicle focuses on how college students' psychological distress accumulates, interacts, and evolves across a semester.

CPCD-Bench

CPCD-Bench evaluates long-horizon campus counseling capabilities from three dimensions:

Task	Description
Session-level Response	Generate appropriate counselor responses using current context and historical memory
Memory Recall	Recall factual information from long counseling histories
Temporal-Causal Reasoning	Analyze chronological event development and causal relationships

Usage

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_name = "EdwinUstb/CPCD-Chat-4B"

tokenizer = AutoTokenizer.from_pretrained(
    model_name,
    trust_remote_code=True
)

model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.bfloat16,
    device_map="auto",
    trust_remote_code=True
)

messages = [
    {
        "role": "user",
        "content": "我最近因为学业和家庭压力感到很焦虑，不知道该怎么办。"
    }
]

text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True
)

inputs = tokenizer([text], return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=512,
    temperature=0.7,
    top_p=0.9
)

response = tokenizer.decode(
    outputs[0][inputs.input_ids.shape[-1]:],
    skip_special_tokens=True
)

print(response)

Intended Use

This model is intended for research on:

• psychological counseling dialogue modeling;
• long-horizon dialogue generation;
• cross-session memory modeling;
• campus mental-health support datasets;
• temporal-causal reasoning in counseling scenarios.

Limitations

CPCD-Chat-8B is trained on synthetic counseling data. It may generate responses that are incomplete, overly generic, or inappropriate in high-risk mental-health situations.

The model should not be used as a substitute for professional psychological counseling, clinical diagnosis, or treatment.

Ethical Considerations

This model is released for research and evaluation purposes only.

Users should be aware that:

• the training data are synthetic and do not represent real counseling records;
• the model may fail to detect or properly handle crisis situations;
• any deployment-oriented use should include professional review, safety monitoring, and clear user-facing disclaimers.

Citation

If you find this model or project useful, please cite:

@misc{gou2026psychronicle,
  title  = {Psy-Chronicle: A Structured Pipeline for Synthesizing Long-Horizon Campus Psychological Counseling Dialogues},
  author = {Chaogui Gou and Jiarui Liang},
  year   = {2026},
  note   = {Preprint},
  url    = {https://github.com/EdwinUSTB/Psy-Chronicle}
}

License

This model is released under the Apache License 2.0.