MVR-Bench: Minimum Viable Relationships Market-Permission Reasoning Benchmark

MVR-Bench is a benchmark dataset for evaluating AI agents on market-permission reasoning in African, emerging, and high-context markets. It was published by African Market OS and originated by Farouk Mark Mukiibi, with a public development split released on May 25, 2026. The benchmark measures the 'Reckless-GO Rate,' which is the share of cases where an AI agent overclaims market-entry readiness.

Use Cases

• Evaluating AI agent recommendations for market-entry readiness based on the described 'Reckless-GO Rate' metric.
• Testing AI systems for overclaiming funding or scaling readiness in African and emerging markets as described.
• Benchmarking AI performance on concepts of local legitimacy, guardian approval, and evidence sufficiency mentioned in the framework.
• Integrating and testing AI agents or Model Context Protocols (MCPs) with market-trust reasoning capabilities as referenced in the public materials.

Strengths

• Includes a defined headline metric ('Reckless-GO Rate') for quantitative evaluation.
• Public materials include a task definition, schemas, a labeled public development split, and methodology notes.
• Has a clear citation anchor with multiple canonical links, including a DOI (10.6084/m9.figshare.30391393).

Limitations

• The private test set, private labels, and core scoring engine are not published, limiting independent full evaluation.
• Row count and column-level documentation are unknown, which may limit suitability assessment.
• The public split is explicitly not licensed for AI/ML training, fine-tuning, or commercial use without authorization.

Provenance

Source

African Market OS, originated by Farouk Mark Mukiibi.

Collection Method

Benchmark dataset created for evaluating AI agent reasoning; methodology notes are included in public materials.

Freshness

Last updated 2026-05-25 13:35:25; freshness should be verified.

Geography

Focus on African, emerging, and high-context markets.