Lightweight LLMs in Mental Healthcare: Systematic Review of Federated Learning, Compression, and Privacy Enhancement
Abstract
Use of mental health and language models involve computational inefficiency, privacy violation, and personalization standards. Our systematic search includes the men addressed issues through a privacy-protecting, lightweight language. models evaluation placing particular focus on federated learning, compression schemes, and personalization techniques. As per PRISMA guidelines, systematically we reviewed 89 Academic articles after January 2019 through September 2025. Our findings prove that privacy-preserving schemes would attain 9095 % precision and reduce the communication cost by$60-80 \%$. Also, many compression schemes attain model size compression by 5-10x with minimal precision loss. Importantly, federated personalization techniques attained a${1 5}-{2 5} \%$improvement in user specific task. We also show that federated learning coupled with lightweight models offers a promising solution to privacyenhancing mental healthcare solutions. A lot remains to be conducted in terms of research, e.g., standardized evaluation metrics by necessity, better privacy guarantees, and clinical validation. This review captures accrued knowledge and provides a research road map with future implications by pointing out how the considered technologies hold promise in transforming mental health support.
How strong is the evidence?
Evidence Hub reads this as a systematic review. Its evidence level is 1 of 5 (Highest · pooled evidence). Pooled evidence from several studies sits at the top of the evidence pyramid.
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.