DiagBench: Multi-Turn Diagnostic Reasoning Benchmark

DiagBench is a multi-center benchmark designed to assess multi-turn diagnostic capabilities across diverse clinical settings. It evaluates an agent's ability to manage an entire diagnostic trajectory from initial patient presentation to final diagnosis by actively selecting examinations. The dataset was created by Henrychur and was last updated on February 12, 2026.

Use Cases

• Benchmarking AI diagnostic agents based on multi-turn clinical dialogue trajectories.
• Training models for sequential clinical decision-making based on active examination selection.
• Evaluating the robustness of diagnostic reasoning across diverse clinical settings mentioned in the description.

Strengths

• The benchmark is multi-center, suggesting data from diverse clinical sources.
• It focuses on multi-turn diagnostic trajectories, a more complex task than single-shot diagnosis.
• It was last updated on February 12, 2026, indicating recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and dataset size are unknown, which may limit suitability assessment.
• The description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Collection Method

Likely compiled for benchmarking purposes from multiple clinical sources.

Freshness

Last updated 2026-02-12 07:44:30.