Psy-Chronicle:A Structured Pipeline for Synthesizing Long-Horizon Campus Psychological Counseling Dialogues

2026-05-22 · Source: cs.CL updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Advanced, extended

Summary

Psy-Chronicle is a structured data-generation framework designed to synthesize long-horizon campus psychological counseling dialogues. It addresses the limitation of existing short-turn data by modeling how college students' psychological distress accumulates and evolves over time. The framework constructs a four-dimensional student profile, generates a semester-spanning temporal stress event graph, and simulates cross-session counseling with structured memory integration. Using Psy-Chronicle, the authors created CPCD, a Chinese long-horizon dialogue dataset containing 100 student profiles, 90,000 counseling dialogues, and approximately 11.45 million Chinese characters. They also developed CPCD-Bench to evaluate models on session-level response, long-horizon memory recall, and temporal-causal reasoning. Experiments show CPCD effectively improves session-level response and memory recall for models like Qwen3-8B, but temporal-causal reasoning remains a significant challenge.

Key takeaway

For AI Scientists and Machine Learning Engineers developing psychological support systems, this research highlights the critical need for long-horizon, context-aware training data. Your models will significantly improve session-level response generation and memory recall by incorporating structured student profiles, temporal stress event graphs, and cross-session memory mechanisms. However, be aware that temporal-causal reasoning remains a key challenge, requiring further research into explicit event-chain annotations or graph-structured reasoning to enhance understanding of distress evolution.

Key insights

Long-horizon psychological counseling data can be synthesized by modeling student profiles, temporal stress events, and cross-session memory.

Principles

• Psychological distress accumulates over time.
• Structured memory aids cross-session continuity.
• Temporal event graphs model stress evolution.

Method

Psy-Chronicle constructs student profiles, generates temporal stress event graphs, simulates interactive counseling between student and counselor agents, and integrates structured memory summaries for continuity.

In practice

• Synthesize long-horizon dialogue datasets.
• Evaluate models on memory recall.
• Train models for cross-session continuity.

Topics

• Psychological Counseling
• Long-Horizon Dialogues
• LLM Data Synthesis
• Temporal Event Graphs
• Student Mental Health
• CPCD Dataset
• Memory Recall

Code references

• EdwinUSTB/Psy-Chronicle

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, NLP Engineer

Related on AIssential

• Tree-of-Experience: A Structured Experience-Management Solution for Self-Evolving Agents under Low-Repetition and Implicit-Reward Environments — Structured experience management is crucial for LLM agents in low-repetition, implicit-reward environments like financial markets.
• Who Am I? History-Aware Profiles for Student Simulation in Tutoring Dialogues — History-conditioned student simulation with RL-trained profiles significantly improves LLM-powered tutoring tool evaluation by predicting student turns.
• ChronoMedKG: A Temporally-Grounded Biomedical Knowledge Graph and Benchmark for Clinical Reasoning — Temporal information is critical for accurate clinical reasoning, a gap ChronoMedKG addresses with its temporally-grounded biomedical knowledge graph.
• Auditing Framing-Sensitive Behavioral Instability in Large Language Models for Mental Health Interactions — LLMs exhibit framing-sensitive behavioral instability in mental health contexts, rooted in internal representations.
• EngramaBench: Evaluating Long-Term Conversational Memory with Structured Graph Retrieval — Structured memory excels at cross-space reasoning, but full-context prompting currently leads in overall performance.
• When Clients Stop Following: A Cognitive Conceptualization Diagram-driven Framework for Strategic Counseling — LLM counseling benchmarks are flawed by client over-cooperation; resistance-aware training and evaluation are crucial for realistic therapeutic progress.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CL updates on arXiv.org.