SciGroveBeta
Medicine

Mental-R1: Aligning LLM Reasoning for Mental Health Assessment

Xin Wang, Boyan Gao, Yibo Yang, David A. Clifton

Featured June 24, 2026

Original Paper

This analysis was generated by SciGrove. Upload your own PDFs or enter a DOI — and get the same AI breakdown on any paper.

Get started

AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content. Always refer to the original paper.

Simply

By teaching large language models to think like people, starting with open exploration and then narrowing down to a confident conclusion across specific mental assessment steps, the paper helps them better understand and assess mental health situations.

In depth
The paper introduces Cognitive Relative Policy Optimization (CRPO), a reinforcement learning framework that aligns large language model (LLM) reasoning with human cognitive processes for mental health assessment. It achieves this by integrating stage-wise entropy regularization to mimic the human uncertainty-to-certainty transition and formalizing theory-grounded cognitive reasoning stages inspired by cognitive appraisal theory. This approach leads to more reliable and interpretable mental health assessments by LLMs.

Key Takeaways

  • 1
    The proposed Cognitive Relative Policy Optimization (CRPO) framework significantly improves LLM performance in mental health assessment by aligning reasoning with human cognitive dynamics.
  • 2
    A novel stage-wise entropy regularization mechanism guides LLMs from broad exploration in early reasoning stages to confident decision-making in later stages, mirroring human cognition.
  • 3
    The framework incorporates theory-grounded cognitive reasoning stages (Stimulus, Primary Appraisal, Secondary Appraisal, Reaction, Mental State) and a balanced reward system to enhance interpretability and address data imbalance.

Conceptual Flow

HIGH LEVEL
1
Methodology: Human-like Reasoning for LLMs

The new method teaches computer models to think step-by-step, just like a person would, starting broad and then getting more certain, to understand feelings better.

Person's Statement
Think Step-by-Step
Mental State Assessment
2
Results: Better Mental Health Assessment

The computer model, trained with this new thinking process, became much better at figuring out mental health conditions compared to other smart computer programs.

Old Computer Models
Less Accurate
New Computer Model
More Accurate