MANDOS: Machine-graded AI Navigation of Dangerous Operational Scenarios

MANDOS is a physics-grounded, recall-sanitized process safety benchmark for AI language models. It tests models against documented human organizational failure modes in high-consequence industrial abnormal operations. The dataset was created by relayforge-ai and last updated on June 4, 2026.

Use Cases

• Benchmarking AI model safety in industrial operations based on the described decision-tree structure.
• Studying organizational failure modes in high-consequence scenarios as referenced in the description.
• Evaluating model recall and reasoning on physics-grounded safety problems as mentioned in the dataset's purpose.

Strengths

• Dataset is specifically designed as a physics-grounded benchmark for AI safety.
• Focuses on documented human organizational failure modes in high-consequence operations.

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

relayforge-ai on Hugging Face

Collection Method

Likely constructed as a decision-tree benchmark from documented case studies.

Freshness

Last updated 2026-06-04 03:02:58; freshness should be verified.