BuddyBench: A Privacy-Constrained Multi-Task Benchmark for Pediatric Social-Communication Personalization

2026-05-27

AI Summary

This work addresses the lack of privacy-preserving, multi-task benchmarks in pediatric neurodevelopmental research that integrate fine-grained learning trajectories, clinical assessments, and intervention outcomes. To bridge this gap, the authors introduce BuddyBench—the first multi-task framework unifying observational cohort data with randomized controlled trials. BuddyBench jointly models drill-level behavioral logs, standardized clinical evaluations, self-reported information, and treatment endpoints, while adhering to strict privacy constraints through both real and synthetic data (BuddyBench-Sim). The benchmark supports four core tasks: knowledge tracing, personalized recommendation, clinical prediction, and causal inference. Experimental results demonstrate that baseline models capture meaningful signals across all tasks, confirming the association between behavioral trajectories and clinical outcomes, thereby establishing a reproducible and privacy-safe paradigm for child development research.

Abstract

BuddyBench introduces a privacy-constrained multi-task benchmark for pediatric social-communication personalization. Unlike existing neurodevelopmental repositories that primarily emphasize imaging, genetics, or cross-sectional clinical phenotyping, BuddyBench links drill-level learning trajectories, standardized clinical assessments, BuddyPlan self-report, and randomized-treatment endpoints within a unified benchmark schema. BuddyBench combines two cohorts: ND-03 is an observational cohort with dense drill coverage for Tasks1-2 (n = 189), and ND-02 is a randomized controlled trial cohort for Tasks3-4 (n = 86 ITT). Together, they support knowledge tracing, next-drill recommendation, clinical prediction, and causal inference, linking behavioral personalization to clinical evaluation. We additionally introduce BuddyBench-Sim, a synthetic companion dataset for reproducible evaluation. Baselines show signal across tasks while keeping pediatric clinical records protected.

Problem

• privacy-constrained
• multi-task benchmark
• pediatric social-communication
• personalization
• clinical evaluation

Innovation

• privacy-constrained benchmark
• multi-task learning
• pediatric personalization
• knowledge tracing
• synthetic dataset