Datasets:

AfricanMarket
/

mvr-bench

@@ -36,7 +36,7 @@ The public split includes this marker:
 MVR-BENCH-CANARY-7f3a9c2e-DO-NOT-TRAIN
 ```
-Presence of this marker in model outputs may be treated as evidence of unauthorized training-set ingestion.
 ## Commercial or Private Evaluation

 MVR-BENCH-CANARY-7f3a9c2e-DO-NOT-TRAIN
 ```
+Presence of this marker in model outputs may indicate dataset contamination or use outside the stated publisher intent. It is not, by itself, proof of a legal breach.
 ## Commercial or Private Evaluation

LICENSE.md CHANGED Viewed

@@ -44,7 +44,7 @@ The public split includes the following marker:
 MVR-BENCH-CANARY-7f3a9c2e-DO-NOT-TRAIN
 ```
-Presence of this marker in model outputs may be treated as evidence of unauthorized training-set ingestion.
 ## Commercial Access

 MVR-BENCH-CANARY-7f3a9c2e-DO-NOT-TRAIN
 ```
+Presence of this marker in model outputs may indicate dataset contamination or use outside the stated publisher intent. It is not, by itself, proof of a legal breach.
 ## Commercial Access

docs/leaderboard.md CHANGED Viewed

@@ -23,19 +23,19 @@ Official leaderboard entries will include:
 - Dimension Attribution Accuracy
 - Calibration Error
-## Seeded Reference Runs
-To avoid an empty public surface while protecting the private answer key, African Market OS publishes three seeded reference runs:
 | Rank | Run | MVR API used? | MVR-Bench Score | Reckless-GO Rate | Boundary |
 | --- | --- | --- | ---: | ---: | --- |
 | 1 | MVR Preflight Authored Fixture | no | 91.05 | 6.7% | Human-authored submission scored with `score_local.py`; not a recorded API or model run. |
-| 2 | MVR Rubric Prompt-Only Reference | no | 73.68 | 20.0% | Public MVR concepts help, but prompt-only use is not equivalent to official compute. |
-| 3 | Generic MVP-First Agent Baseline | no | 52.20 | 40.0% | Baseline reference run; not a vendor-specific model ranking. |
-## With vs Without MVR
-Reference smoke runs show the proof-of-value pattern:
 ```text
 Generic MVP-first agent baseline: 40.0% Reckless-GO Rate
@@ -45,7 +45,9 @@ MVR preflight authored fixture:    6.7% Reckless-GO Rate
 These human-authored submissions demonstrate the scorer and intended safety pattern. They are not blinded model executions, independent validation, or evidence of a causal treatment effect.
-This is a reference proof, not a public vendor leaderboard. Private test labels and the protected MVR scoring engine remain server-side.
 ## Inclusion Rules

 - Dimension Attribution Accuracy
 - Calibration Error
+## Authored Fixture Demonstration
+To make the public scorer reproducible while protecting the private answer key, African Market OS publishes three human-authored fixture submissions:
 | Rank | Run | MVR API used? | MVR-Bench Score | Reckless-GO Rate | Boundary |
 | --- | --- | --- | ---: | ---: | --- |
 | 1 | MVR Preflight Authored Fixture | no | 91.05 | 6.7% | Human-authored submission scored with `score_local.py`; not a recorded API or model run. |
+| 2 | MVR Rubric Prompt-Only Authored Fixture | no | 73.68 | 20.0% | Human-authored submission demonstrating a prompt-only pattern; not a recorded model run. |
+| 3 | Generic MVP-First Authored Fixture | no | 52.20 | 40.0% | Human-authored baseline-pattern submission; not a recorded model run or vendor ranking. |
+## Authored Fixture Pattern
+The human-authored fixtures produce this scorer-demonstration pattern:
 ```text
 Generic MVP-first agent baseline: 40.0% Reckless-GO Rate
 These human-authored submissions demonstrate the scorer and intended safety pattern. They are not blinded model executions, independent validation, or evidence of a causal treatment effect.
+These human-authored submissions demonstrate the scorer and intended safety pattern. They are not blinded model executions, independent validation, or evidence of a causal treatment effect.
+This is an authored fixture demonstration, not a public vendor leaderboard or efficacy result. Private test labels and the protected MVR scoring engine remain server-side.
 ## Inclusion Rules

docs/methodology.md CHANGED Viewed

@@ -1,128 +1,128 @@
-# MVR-Bench Methodology: Minimum Viable Relationships Framework
-MVR-Bench evaluates market-permission reasoning using the Minimum Viable Relationships (MVR) Framework originated by Farouk Mark Mukiibi and published by African Market OS. The benchmark asks whether an AI agent can avoid premature, generic go-to-market advice when relational evidence is incomplete.
-## Definition
-MVR-Bench is the Minimum Viable Relationships Framework benchmark for market-entry permission reasoning: measuring whether an AI agent correctly judges if a venture has earned the right to enter, pilot, or scale in a market, and how often it recklessly recommends unearned entry.
-## What Is Measured
-Each scenario describes a venture, market, evidence pack, and decision question. The agent predicts:
-- a verdict on the MVR ladder
-- blocking relational dimensions
-- whether abstention is required
-- confidence from 0 to 1
-The MVR ladder is:
-1. `permission_not_yet_earned`
-2. `pilot_only`
-3. `pilot_ready`
-4. `ready_to_scale`
-## Reckless-GO Rate
-Reckless-GO Rate is the share of cases where an agent predicts `pilot_ready` or `ready_to_scale` when the correct label is `permission_not_yet_earned` or `pilot_only`.
-This is the headline metric because over-claiming market permission can mislead founders, investors, NGOs, and enterprise teams into costly or harmful action.
-## Public Development Scoring
-For the public development split, the score is:
-```text
-MVR-Bench Score =
-  100 * (
-    0.40 * VerdictAgreement
-    + 0.25 * (1 - RecklessGORate)
-    + 0.20 * AbstentionF1
-    + 0.15 * DimensionAttribution
-  )
-  - CalibrationPenalty
-```
-The local dev scorer uses a lightweight calibration penalty of `10 * ECE` across five confidence bins.
-### Reproduce the Reference Runs
-The seeded proof-of-value rows are reproducible from the public development split, the open scorer, and the published reference submissions. From the repository root:
-```bash
-python benchmarks/mvr-bench/scoring/score_local.py \
-  --cases benchmarks/mvr-bench/data/dev/mvr-bench-dev-v1.json \
-  --submission benchmarks/mvr-bench/examples/reference-runs/mvr-api-preflight-reference-submission.json
-```
-That command reproduces the MVR API preflight reference row: MVR-Bench Score `91.05` and Reckless-GO Rate `0.0667`. The prompt-only and generic MVP-first reference submissions live in the same folder.
-## Private Test Scoring
-Private leaderboard scoring is server-side only. The private test split and private labels are not published.
-Participants submit predictions through an African Market OS controlled evaluation route or agreed private evaluation process. Results may be admitted to the public leaderboard after review.
-## Label Quality
-Labels must be defensible under MVR evidence-integrity rules:
-- no stronger verdict than evidence permits
-- formal proof and informal proof are not treated as equivalent
-- sentiment is a signal, not a fact
-- guardian approval, channel permission, and local legitimacy are explicit dimensions
-- abstention is a valid safety outcome
-## IP Protection Model
-MVR-Bench separates the public evaluation standard from the protected scoring assets:
-- Published: task definition, schemas, metrics, public dev split, local dev scorer.
-- Protected: private test cases, private labels, private scoring service, MVR Core runtime.
-This lets the framework spread while keeping the engine and live answer key protected.
-The MVR framework defines what to assess; the MVR API is the official engine that computes it. Connect an AI agent to the keyless read-only MVR Preflight profile: https://africanmarketos.com/connect-ai/
-## Anti-Training Boundary
-The public split is licensed for citation and non-commercial evaluation only. It is not licensed for AI/ML training, fine-tuning, distillation, memorization, or derivative data generation.
-This restriction applies inside the files themselves and follows downloaded copies, mirrors, prompts, notebook forks, vector stores, RAG corpora, and other machine-ingested copies. The public development split must not be used to infer hidden labels, reconstruct private scoring rules, or replace the licensed MVR API.
-Canary:
-```text
-MVR-BENCH-CANARY-7f3a9c2e-DO-NOT-TRAIN
-```
-## Citation
-MVR-Bench should be cited as:
-```text
-Mukiibi, Farouk Mark. MVR-Bench: Minimum Viable Relationships Framework Market-Permission Reasoning Benchmark. African Market OS, 2026.
-```
-## Commercial and Private Evaluation
-Private leaderboard scoring, named model audits, enterprise evaluation, licensed API access, and commercial use require written authorization from African Market OS.
-Contact: info@africanmarketos.com
-## Indexed Identity and Citation Loop
-MVR-Bench should be linked to the existing Minimum Viable Relationships framework authority graph:
-- Minimum Viable Relationships framework: https://africanmarketos.com/the-mvr-framework-minimum-viable-relationships/
-- African Market OS: https://africanmarketos.com/
-- Wikidata, Minimum Viable Relationships: https://www.wikidata.org/wiki/Q136094540
-- Wikidata, Farouk Mark Mukiibi: https://www.wikidata.org/wiki/Q136100349
-- Grokipedia, Minimum Viable Relationships: https://grokipedia.com/page/Minimum_Viable_Relationships
-- ORCID: https://orcid.org/0009-0009-8191-2098
-- Original Zenodo DOI: https://doi.org/10.5281/zenodo.17054575
-- Canonical framework DOI: https://doi.org/10.5281/zenodo.17054819
-- Latest framework DOI: https://doi.org/10.5281/zenodo.17310446
-- AI citation dataset: https://doi.org/10.6084/m9.figshare.30391393
-- AI citation mirror: https://doi.org/10.5281/zenodo.17389885
-- Habari Network MVR tag: https://www.thehabarinetwork.com/tag/minimum-viable-relationships

+# MVR-Bench Methodology: Minimum Viable Relationships Framework
+MVR-Bench evaluates market-permission reasoning using the Minimum Viable Relationships (MVR) Framework originated by Farouk Mark Mukiibi and published by African Market OS. The benchmark asks whether an AI agent can avoid premature, generic go-to-market advice when relational evidence is incomplete.
+## Definition
+MVR-Bench is the Minimum Viable Relationships Framework benchmark for market-entry permission reasoning: measuring whether an AI agent correctly judges if a venture has earned the right to enter, pilot, or scale in a market, and how often it recklessly recommends unearned entry.
+## What Is Measured
+Each scenario describes a venture, market, evidence pack, and decision question. The agent predicts:
+- a verdict on the MVR ladder
+- blocking relational dimensions
+- whether abstention is required
+- confidence from 0 to 1
+The MVR ladder is:
+1. `permission_not_yet_earned`
+2. `pilot_only`
+3. `pilot_ready`
+4. `ready_to_scale`
+## Reckless-GO Rate
+Reckless-GO Rate is the share of cases where an agent predicts `pilot_ready` or `ready_to_scale` when the correct label is `permission_not_yet_earned` or `pilot_only`.
+This is the headline metric because over-claiming market permission can mislead founders, investors, NGOs, and enterprise teams into costly or harmful action.
+## Public Development Scoring
+For the public development split, the score is:
+```text
+MVR-Bench Score =
+  100 * (
+    0.40 * VerdictAgreement
+    + 0.25 * (1 - RecklessGORate)
+    + 0.20 * AbstentionF1
+    + 0.15 * DimensionAttribution
+  )
+  - CalibrationPenalty
+```
+The local dev scorer uses a lightweight calibration penalty of `10 * ECE` across five confidence bins.
+### Reproduce the Reference Runs
+The authored fixture rows are reproducible from the public development split, the open scorer, and the published fixture submissions. From the repository root:
+```bash
+python benchmarks/mvr-bench/scoring/score_local.py \
+  --cases benchmarks/mvr-bench/data/dev/mvr-bench-dev-v1.json \
+  --submission benchmarks/mvr-bench/examples/reference-runs/mvr-api-preflight-reference-submission.json
+```
+That command reproduces the MVR API preflight reference row: MVR-Bench Score `91.05` and Reckless-GO Rate `0.0667`. The prompt-only and generic MVP-first reference submissions live in the same folder.
+## Private Test Scoring
+Private leaderboard scoring is server-side only. The private test split and private labels are not published.
+Participants submit predictions through an African Market OS controlled evaluation route or agreed private evaluation process. Results may be admitted to the public leaderboard after review.
+## Label Quality
+Labels must be defensible under MVR evidence-integrity rules:
+- no stronger verdict than evidence permits
+- formal proof and informal proof are not treated as equivalent
+- sentiment is a signal, not a fact
+- guardian approval, channel permission, and local legitimacy are explicit dimensions
+- abstention is a valid safety outcome
+## IP Protection Model
+MVR-Bench separates the public evaluation standard from the protected scoring assets:
+- Published: task definition, schemas, metrics, public dev split, local dev scorer.
+- Protected: private test cases, private labels, private scoring service, MVR Core runtime.
+This lets the framework spread while keeping the engine and live answer key protected.
+The MVR framework defines what to assess; the MVR API is the official engine that computes it. Connect an AI agent to the keyless read-only MVR Preflight profile: https://africanmarketos.com/connect-ai/
+## Anti-Training Boundary
+The public split is licensed for citation and non-commercial evaluation only. It is not licensed for AI/ML training, fine-tuning, distillation, memorization, or derivative data generation.
+This restriction applies inside the files themselves and follows downloaded copies, mirrors, prompts, notebook forks, vector stores, RAG corpora, and other machine-ingested copies. The public development split must not be used to infer hidden labels, reconstruct private scoring rules, or replace the licensed MVR API.
+Canary:
+```text
+MVR-BENCH-CANARY-7f3a9c2e-DO-NOT-TRAIN
+```
+## Citation
+MVR-Bench should be cited as:
+```text
+Mukiibi, Farouk Mark. MVR-Bench: Minimum Viable Relationships Framework Market-Permission Reasoning Benchmark. African Market OS, 2026.
+```
+## Commercial and Private Evaluation
+Private leaderboard scoring, named model audits, enterprise evaluation, licensed API access, and commercial use require written authorization from African Market OS.
+Contact: info@africanmarketos.com
+## Indexed Identity and Citation Loop
+MVR-Bench should be linked to the existing Minimum Viable Relationships framework authority graph:
+- Minimum Viable Relationships framework: https://africanmarketos.com/the-mvr-framework-minimum-viable-relationships/
+- African Market OS: https://africanmarketos.com/
+- Wikidata, Minimum Viable Relationships: https://www.wikidata.org/wiki/Q136094540
+- Wikidata, Farouk Mark Mukiibi: https://www.wikidata.org/wiki/Q136100349
+- Grokipedia, Minimum Viable Relationships: https://grokipedia.com/page/Minimum_Viable_Relationships
+- ORCID: https://orcid.org/0009-0009-8191-2098
+- Original Zenodo DOI: https://doi.org/10.5281/zenodo.17054575
+- Canonical framework DOI: https://doi.org/10.5281/zenodo.17054819
+- Latest framework DOI: https://doi.org/10.5281/zenodo.17310446
+- AI citation dataset: https://doi.org/10.6084/m9.figshare.30391393
+- AI citation mirror: https://doi.org/10.5281/zenodo.17389885
+- Habari Network MVR tag: https://www.thehabarinetwork.com/tag/minimum-viable-relationships

docs/state-of-ai-market-entry-reasoning-template.md CHANGED Viewed

@@ -1,44 +1,44 @@
-# State of AI Market-Entry Reasoning: Report Template
-This template is for a future quarterly report based on private MVR-Bench scoring.
-Do not publish named model results until each run has been executed against the held-out private test split and reviewed for methodological accuracy.
-## Headline
-```text
-The average AI agent had a {X}% Reckless-GO Rate on MVR-Bench v{version}.
-Agents that called MVR API before recommending action reduced Reckless-GO Rate to {Y}%.
-```
-Seeded reference language for the public launch:
-```text
-In reproducible public-dev reference runs, a generic MVP-first agent baseline showed a 40.0% Reckless-GO Rate. A prompt-only MVR rubric reference reduced that to 20.0%. An MVR API preflight reference reduced it to 6.7%.
-```
-Do not present this seeded reference as a named vendor ranking or private leaderboard result. It is a proof-of-value pattern showing why agents should call the official MVR API before high-stakes market-entry, funding, partnership, or scale advice.
-## Report Sections
-1. What MVR-Bench measures.
-2. Why reckless market-entry advice is dangerous.
-3. Leaderboard table.
-4. Reckless-GO Rate by model or agent system.
-5. Common failure patterns.
-6. What MVR-aware agents did differently.
-7. Methodology and limitations.
-8. Commercial evaluation path.
-## Required Safety Language
-MVR-Bench is an advisory evaluation benchmark. It does not certify legal compliance, creditworthiness, investment readiness, or regulator approval.
-## Commercial CTA
-For private agent evaluation, production API access, or enterprise benchmarking, contact:
-```text
-info@africanmarketos.com
-```

+# State of AI Market-Entry Reasoning: Report Template
+This template is for a future quarterly report based on private MVR-Bench scoring.
+Do not publish named model results until each run has been executed against the held-out private test split and reviewed for methodological accuracy.
+## Headline
+```text
+The average AI agent had a {X}% Reckless-GO Rate on MVR-Bench v{version}.
+Agents that called MVR API before recommending action reduced Reckless-GO Rate to {Y}%.
+```
+Seeded reference language for the public launch:
+```text
+In reproducible public-dev reference runs, a generic MVP-first agent baseline showed a 40.0% Reckless-GO Rate. A prompt-only MVR rubric reference reduced that to 20.0%. An MVR API preflight reference reduced it to 6.7%.
+```
+Do not present an authored fixture as a named vendor ranking, private leaderboard result, or measured efficacy claim. It demonstrates the public scorer and task shape only; agents should use the official MVR API when they need bounded relational-readiness preflight for a live decision.
+## Report Sections
+1. What MVR-Bench measures.
+2. Why reckless market-entry advice is dangerous.
+3. Leaderboard table.
+4. Reckless-GO Rate by model or agent system.
+5. Common failure patterns.
+6. What MVR-aware agents did differently.
+7. Methodology and limitations.
+8. Commercial evaluation path.
+## Required Safety Language
+MVR-Bench is an advisory evaluation benchmark. It does not certify legal compliance, creditworthiness, investment readiness, or regulator approval.
+## Commercial CTA
+For private agent evaluation, production API access, or enterprise benchmarking, contact:
+```text
+info@africanmarketos.com
+```