# Reliable Self-Harm Risk Screening via Adaptive Multi-Agent LLM Systems

arXiv: https://arxiv.org/abs/2604.22154 · HTML: https://mlsanity.com/paper/2604.22154/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Medical and scientific domain agents, one of 16 directions in LLM agents and tool use.
- Builds on 9 of its 25 references that are in the map.
- First posted on arXiv 24 Apr 2026.

## Placed in

- LLM agents and tool use / [Medical and scientific domain agents](https://mlsanity.com/agents/medical-and-scientific-domain-agents/)

## Builds on (most cited first, 9 of 9)

- [Ethical and social risks of harm from Language Models](https://mlsanity.com/paper/2112.04359/) (2021)
- [ToxicChat: Unveiling Hidden Challenges of Toxicity Detection in Real-World User-AI Conversation](https://mlsanity.com/paper/2310.17389/) (2023)
- [MentalBERT: Publicly Available Pretrained Language Models for Mental Healthcare](https://mlsanity.com/paper/2110.15621/) (2021)
- [Aegis2.0: A Diverse AI Safety Dataset and Risks Taxonomy for Alignment of LLM Guardrails](https://mlsanity.com/paper/2501.09004/) (2025)
- [AgentNet: Decentralized Evolutionary Coordination for LLM-based Multi-Agent Systems](https://mlsanity.com/paper/2504.00587/) (2025)
- [Depression and Self-Harm Risk Assessment in Online Forums](https://mlsanity.com/paper/1709.01848/) (2017)
- [ScAN: Suicide Attempt and Ideation Events Dataset](https://mlsanity.com/paper/2205.07872/) (2022)
- [Detecting Suicidal Ideation in Chinese Microblogs with Psychological Lexicons](https://mlsanity.com/paper/1411.0778/) (2014)
- [Constrained Process Maps for Multi-Agent Generative AI Workflows](https://mlsanity.com/paper/2602.02034/) (2026)

## Abstract (arXiv)

Emerging AI systems in behavioral health and psychiatry use multi-step or multi-agent LLM pipelines for tasks like assessing self-harm risk and screening for depression. However, common evaluation approaches, like LLM-as-a-judge, do not indicate when a decision is reliable or how errors may accumulate across multiple LLM judgements, limiting their suitability for safety-critical settings. We present a statistical framework for multi-agent pipelines structured as directed acyclic graphs (DAGs) that provides an alternative to heuristic voting with principled, adaptive decision-making. We model each agent as a stochastic categorical decision and introduce (1) tighter agent-level performance confidence bounds, (2) a bandit-based adaptive sampling strategy based on input difficulty, and (3) regret guarantees over the multi-agent system that shows logarithmic error growth when deployed. We evaluate our system on two labeled datasets in behavioral health : the AEGIS 2.0 behavioral health subset (N=161) and a stratified sample of SWMH Reddit posts (N=250). Empirically, our adaptive sampling strategy achieves the lowest false positive rate of any condition across both datasets, 0.095 on AEGIS 2.0 compared to 0.159 for single-agent models, reducing incorrect flagging of safe content by 40% and still having similar false negative rates across all conditions. These results suggest that principled adaptive sampling offers a meaningful improvement in precision without reducing recall in this setting.