MedMemoryBench: Bilingual Benchmark for Long-Term Memory in Healthcare Agents

MedMemoryBench evaluates AI agents' long-term memory across 12 months of simulated chronic disease management. The benchmark contains 20 diverse patient personas and bilingual Chinese/English dialogues for longitudinal patient-doctor interactions. It was created by Cyan27 and last updated on Hugging Face in May 2026.

Use Cases

• Benchmarking long-term memory retention in AI agents based on multi-session clinical dialogues
• Training healthcare chatbots to manage chronic diseases based on longitudinal patient data
• Evaluating multilingual (Chinese/English) information recall in medical AI systems

Strengths

• Simulates 12 months of longitudinal patient-doctor interactions
• Includes 20 diverse patient personas for varied testing scenarios
• Features bilingual (Chinese/English) text for cross-lingual evaluation

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

Cyan27 on Hugging Face

Collection Method

Likely constructed for benchmarking AI agent memory in healthcare scenarios.

Time Range

Simulates a 12-month period of chronic disease management.

Freshness

Last updated 2026-05-07 09:42:35