CounselPlex improves counseling by tracking stressors in real time
CounselPlex: A Stressor-Aware Full-Duplex Counseling Model
Computation and LanguageSound
Summary
Counseling conversations need the counselor to understand ongoing stressors, not just the latest words spoken. The authors identify a problem called 'Stressor Blindness,' where models forget the seeker's underlying issues during conversation. They created CounselPlex, a system that keeps track of stress signals even during silent moments and uses this information continuously to improve response quality. Tests show this approach helps the model provide better, more grounded counseling.
What this means in practice
- •For telehealth service providers: Deliver real-time counseling with AI that understands ongoing client stress even during pauses.$Commercial implications: Enables telehealth platforms to offer more empathetic AI counselors that maintain context, improving user experience and outcomes.
- •For call center operations: Enhance automated support agents' ability to identify and respond to caller stress across dialog turns.
Authors
Yeonju Kim, Bella Godiva, Yong Man Ro
Abstract
Full-duplex speech models can listen and speak at the same time, enabling natural, low-latency interaction. In counseling, however, these models often focus on the latest utterance and lose track of the seeker's underlying problem. We call this failure Stressor Blindness. We propose CounselPlex, a stressor-aware full-duplex framework that uses two operations. CounselPlex updates a short stressor state during seeker silence and injects this state into <PAD> positions while listening. We also introduce RoPE Position Shift Augmentation (RPSA), which broadens positional coverage during training and improves performance at longer turn depths in our evaluation. LLM-judge evaluations show improved counseling quality and stressor grounding over strong baselines.