Synthetic Longitudinal Health Profiles for AI Assistant Benchmarking

ESL-bench provides virtual health user profiles containing complete health data, event timelines, clinical exam records, and knowledge-graph-grounded queries. The dataset is designed for use with the HolyEval framework to evaluate AI health assistants. It was created by healthmemoryarena and last updated in April 2026.

Use Cases

• Benchmarking AI agent performance on health profile queries grounded in a clinical knowledge graph.
• Evaluating AI assistant responses to synthetic clinical exam data and event timelines from virtual users.
• Training or testing models on longitudinal patient data simulation using complete health profiles.

Strengths

• Contains complete synthetic health profiles integrating multiple data types per virtual user.
• Designed specifically for structured evaluation within the HolyEval framework.

Limitations

• All data is synthetic, limiting real-world applicability and potential for distribution shift.
• Key quantitative details like row count, column count, and sample size are unknown.

Provenance

Source

healthmemoryarena

Collection Method

Synthetically generated for benchmarking.

Freshness

Last updated April 2026.