SenTSR-Bench: A Real-World Multivariate Time-Series Diagnostic Reasoning Benchmark

SenTSR-Bench is a de-identified, real-world multivariate time-series benchmark for diagnostic reasoning. It was released by Zelin He and collaborators as part of a research paper titled 'SenTSR-Bench: Thinking with Injected Knowledge for Time-Series Reasoning'.

Use Cases

• Benchmarking diagnostic reasoning models based on multivariate time-series data
• Evaluating the effect of knowledge injection on time-series reasoning tasks
• Developing models for real-world diagnostic applications based on de-identified time-series data

Strengths

• Data is de-identified, which suggests privacy considerations were addressed
• Described as a real-world benchmark, implying practical relevance
• Associated with a named research paper, providing academic context

Limitations

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

Provenance

Source

Released as part of the paper 'SenTSR-Bench: Thinking with Injected Knowledge for Time-Series Reasoning'.

Collection Method

De-identified real-world data; specific collection method is unknown.

Freshness

Last updated 2026-02-23 23:04:43; freshness should be verified