FactoryBench: A Benchmark for Machine-Behavior Reasoning in Industrial Robotics

FactoryBench is a benchmark dataset for evaluating machine-behavior reasoning in time-series models and LLMs. It contains question-answer pairs organized along the four levels of Pearl's causal hierarchy, from state identification to intervention prediction. The dataset was created by Forgis and was last updated on June 4, 2026.

Use Cases

• Benchmarking time-series models on operational state identification based on raw robotic telemetry signals.
• Evaluating large language models on causal intervention predictions for industrial robotic systems.
• Training models to reason about machine behavior across Pearl's causal hierarchy levels.

Strengths

• Structured around Pearl's causal hierarchy, providing a multi-level evaluation framework.
• Specifically designed for machine-behavior reasoning in industrial robotic telemetry.

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

Forgis on Hugging Face.

Freshness

Last updated 2026-06-04 14:03:03; freshness should be verified.