Tag
This paper investigates why LLMs fail to apply CBT principles effectively despite high theoretical accuracy, introducing a knowledge-guided framework and a behavioral metric (Protocol Leverage Force) to measure intervention shifts. Experiments show that even with multi-chain-of-thought prompting, models remain biased toward validation and reflection.
Researchers from Lanzhou University propose a CBT-grounded framework for LLM-based psychological counseling that addresses the 'counselor-following' phenomenon in existing benchmarks, introducing CARS (a resistant client simulator), STREAMS (a dual-module strategic reasoning framework), and EWTS-MI (an entropy-weighted evaluation metric) to better handle real-world resistant counseling interactions.
DMT-CBT proposes a framework for longitudinal therapeutic state modeling in CBT counseling, addressing the need for multi-session, multimodal inference and intervention. It introduces DMTCorpus, a synthetic multi-session dataset, and shows improvements in counseling fidelity and therapeutic alliance over existing methods.