Multi-task language model improves suicide risk assessment on social media
Explainable Suicide Risk Assessment on Social Media with Multi-Task QLoRA
Computation and LanguageMachine Learning
Summary
Detecting suicide risk on social media is important but challenging. The authors built a computer program that not only judges how risky a post is but also points out exact words supporting that judgment and identifies factors that might increase or decrease the risk. They trained this program on three related tasks and combined different models to improve accuracy. Their system scored well in a public competition, showing that training for each part carefully and merging results smartly helps the program better explain its decisions.
What this means in practice
- •For mental health support teams: Identify suicide risk levels and relevant factors from social media posts to assist in early intervention efforts.
- •For social media monitoring teams: Highlight specific language and protective or risk factors in posts for more transparent and actionable content review.
Authors
Xuan Zhong Feng, Geoffrey Martin, Hexin Dong, Yifan Peng
Abstract
Explainable suicide-risk assessment requires models not only to estimate risk severity, but also to identify supporting language and the risk and protective factors expressed in a post. We present our system for the IEEE BigData 2026 Cup on Explainable Suicide Risk Assessment on Social Media, which addresses three tasks: risk-level classification, evidence phrase extraction, and multi-label factor identification. Our approach adapts Qwen2.5-Instruct models using quantized low-rank adaptation (QLoRA) and an answer-masked causal language-model objective. We jointly train across all three tasks for risk classification, jointly train on Tasks~1a and 1b for evidence extraction, and adapt Task~2 separately for factor identification. We also tailor aggregation to each output: we average risk-level probabilities from the 32B and 72B models, combine evidence phrases through cross-fold consensus, and calibrate factor-specific decisions through rate matching based on out-of-fold operating points. On the official leaderboard, the final system achieved a composite score of 0.7738, with 0.8089 on Task~1 and 0.6919 on Task~2. Across the evaluated configurations, three-task training performed best for Task~1a, joint training on Tasks~1a and 1b performed best for Task~1b, and task-specific training performed best for Task~2. Probability averaging further improved Task~1a when component models had complementary errors. These findings highlight the value of tailoring both training objectives and aggregation strategies to the output structure of each task within a unified language-model framework.