Synthetic dataset
Grounded Commercial Case Reasoning
Evaluate evidence retrieval and structured reasoning over synthetic contracts, construction cases and commercial disputes.
Nairobi Supplied reference · verified locally · Version 1.0.0
- Reasoning items
- 5,535
- Task types
- 5
- Underlying case pairs
- 1,581
- Released format
- JSONL
This is one supplied Nairobi legal appeals reference associated with Predict Dispute Outcomes and Legal Reasoning Benchmark. Retrospective tables, the safe pre-outcome projection and reasoning tasks have distinct uses. These counts do not define other legal products, Lagos or custom orders.
Overview
Connect dispute facts, contracts and outcomes to prototype legal issue extraction and evidence-based reasoning.
The record concepts include case chronology, claim / award outcome, contract context, case timeline, resolution outcome, legal-event history and contract milestones. Trace obligations and relevant project events. Keep conflicting or missing evidence visible. Distinguish recorded facts from allegations and interpretation.
This dataset is prepared to order. Choose the city, date range and required scope; the accepted order defines the population, final fields, formats and delivery.
- Onboarded city options
- Nairobi (Kenya) · Lagos (Nigeria)
- Custom city requests
- African Major Cities · Additional cities require availability, scope and quote confirmation.
- Data type
- Synthetic
- Connected domains
- Legal, Contracts, Construction, Business
- Catalogue structure
- Longitudinal benchmark
Key features
Record concepts for your order-specific extract.
Legal
Case chronology and recorded resolutions
Contracts
Contract circumstances and effective dates
Construction
Project events and development context
Business
Business histories and relationships
Who it is for
- Legal-tech teams
- LLM evaluators
- dispute professionals
- researchers
Data & Schema
- Entities represented
- Case + document
This identifies the entities described by the catalogue. The analytical row grain, eligibility, keys and dates are agreed for your extract.
Explore supported case analysis across contracts, construction events and dispute records.
Record concepts for your extract
- case chronology
- claim / award outcome
- contract context
- case timeline
- resolution outcome
- legal-event history
- contract milestones
The agreed specification defines the final columns, keys, relationships and observation period for your order.
Supplied reference · verified locally
Nairobi Synthetic Legal Disputes & Appeals — V1
This is one supplied Nairobi legal appeals reference associated with Predict Dispute Outcomes and Legal Reasoning Benchmark. Retrospective tables, the safe pre-outcome projection and reasoning tasks have distinct uses. These counts do not define other legal products, Lagos or custom orders.
- Reference release
- SD-LAW-APPEALS-V1
- Release date
- 2026-10-02
- City
- Nairobi
- Source-dispute/appeal pairs
- 1,581
- Reasoning question items
- 5,535
- Reasoning JSONL fields
- 13
- Observation cutoff
- 2026-09-01
- Original export format
- CSV (gzip) + JSONL
| Configuration | Train | Validation | Test | Total |
|---|---|---|---|---|
| case summaryFile partition | 1,581 | — | — | 1,581 |
| case timelineFile partition | 36,456 | — | — | 36,456 |
| appeal staysFile partition | 396 | — | — | 396 |
| appeal labelsFile partition | 1,581 | — | — | 1,581 |
| pre outcome timelineFile partition | 31,457 | — | — | 31,457 |
| reasoning benchmark | 3,858 | 835 | 842 | 5,535 |
The five CSV configurations load as a single HF train split. Actual analytical case membership is in benchmark/case_splits.csv (1,106/237/238); reasoning questions use 3,858/835/842. Keep linked cases and questions in their published split.
HF train is a file-loading packaging label, not assignment of all cases/events to an analytical training cohort.
Question counts follow the underlying case split; they are not case counts.
Analytical benchmark splits
Split unit: Source-dispute/appeal pair
- Train pairs
- 1,106
- Validation pairs
- 237
- Test pairs
- 238
- Reasoning question splits
- 3,858 / 835 / 842 (train / validation / test)
case summary reference schema
Retrospective source-dispute/appeal pair, including outcomes.
Logical types follow the supplied SCHEMA.json and DATA_DICTIONARY.md. SequenceWithinDate is displayed as observed integer where the source dictionary incorrectly declares date. No formal primary/foreign-key or nullable constraints were supplied.
Selected fields · 12 of 99 columns
Full field definitions, types and descriptions 99 columns
Study, context and observation fields
Technical identifiers and evidence references
case timeline reference schema
Retrospective outcome-complete context; not safe prospective model inputs.
Logical types follow the supplied SCHEMA.json and DATA_DICTIONARY.md. SequenceWithinDate is displayed as observed integer where the source dictionary incorrectly declares date. No formal primary/foreign-key or nullable constraints were supplied.
Selected fields · 12 of 88 columns
Full field definitions, types and descriptions 88 columns
Study, context and observation fields
Technical identifiers and evidence references
appeal stays reference schema
Use source/appeal pair keys to join the supplied reference tables.
Logical types follow the supplied SCHEMA.json and DATA_DICTIONARY.md. SequenceWithinDate is displayed as observed integer where the source dictionary incorrectly declares date. No formal primary/foreign-key or nullable constraints were supplied.
Selected fields · 10 of 15 columns
Full field definitions, types and descriptions 15 columns
Study, context and observation fields
Technical identifiers and evidence references
appeal labels reference schema
Use source/appeal pair keys to join the supplied reference tables.
Logical types follow the supplied SCHEMA.json and DATA_DICTIONARY.md. SequenceWithinDate is displayed as observed integer where the source dictionary incorrectly declares date. No formal primary/foreign-key or nullable constraints were supplied.
Selected fields · 12 of 16 columns
Full field definitions, types and descriptions 16 columns
Study, context and observation fields
Technical identifiers and evidence references
pre outcome timeline reference schema
All and only eligible pre-resolution rows; 12 original columns removed and three time-known stay fields added.
Logical types follow the supplied SCHEMA.json and DATA_DICTIONARY.md. SequenceWithinDate is displayed as observed integer where the source dictionary incorrectly declares date. No formal primary/foreign-key or nullable constraints were supplied.
Selected fields · 12 of 79 columns
Full field definitions, types and descriptions 79 columns
Study, context and observation fields
Technical identifiers and evidence references
reasoning benchmark reference schema
Five grounded question tasks with evidence keys/dates. Questions sharing a source or appeal case remain in the case’s benchmark split.
JSON value serialization is observed across the supplied reasoning files; a separate formal logical-type schema is not claimed.
Selected fields · 8 of 13 columns
Full field definitions, types and descriptions 13 columns
Study, context and observation fields
Technical identifiers and evidence references
Reference data files
benchmark/reasoning_questions.{split}.jsonl
Verified reference sample
Genuine Nairobi legal reference reasoning sample
CC-BY-NC-4.0 · Reference sample license
- Synthetic reference rows
- 30
- Fields
- 13
Original supplied sample bytes and recorded manifest hash retained; all 30 records match the corresponding reference files. Public-repository byte parity has not been established.
The original supplied package contained a no-grant licence notice. The publisher selected CC-BY-NC-4.0 for these public reference materials on 2026-10-13; this is a new publisher choice, not a claim about the original notice or current remote repository licence.
- Thirty existing grounded reasoning items from the supplied reference benchmark; not a separately certified marketplace population sample.
- Answers and evidence metadata are evaluation materials; maintain their case-level split membership.
- This sample describes the supplied Nairobi reference, not a custom commercial order.
Reference sample schema preview
Selected fields · 8 of 13 columns
Full field definitions, types and descriptions 13 columns
Study, context and observation fields
Technical identifiers and evidence references
Illustrative format preview
Example columns and rows illustrate the proposed format; they do not verify paid-release records, fields or relationships.
Proposed preview columns
Selected fields · 10 of 11 columns
All proposed preview columns 11 columns
These names describe the illustrative preview. Final fields and definitions are agreed for your order.
Study, context and observation fields
Technical identifiers and evidence references
Use Cases
Published reference applications
- Evidence-grounded appeal disposition question answering
- Source-resolution effect/change reasoning
- Stay decision and scope grounding
- Case-grouped synthetic benchmark evaluation
Custom study design
Use reviewed issue-to-evidence links and reference findings; score retrieval and reasoning separately, without treating a generated analysis as a legal determination.
This is one supplied Nairobi legal appeals reference associated with Predict Dispute Outcomes and Legal Reasoning Benchmark. Retrospective tables, the safe pre-outcome projection and reasoning tasks have distinct uses. These counts do not define other legal products, Lagos or custom orders.
Proposed questions for your custom order
- Can an AI connect contract obligations, construction events and dispute claims into a supported case analysis while identifying conflicting or missing evidence?
- Can AI separate the parties, allegations and recorded outcome of a case?
- Does an AI assistant retrieve the clause or case record relevant to a question?
- Can an AI assistant distinguish a documented fact from a proposed interpretation?
- Does an AI assistant avoid inventing legal authorities or missing case facts?
Proposed research and analytical applications
- Task prototyping
- Evidence tracing
- Error analysis
- Controlled evaluation
Study and modeling guidance
- Grounded reference tasks
- The 5,535 JSONL questions cover appeal disposition, source-resolution effect/change and stay decision/scope. Answers resolve to supplied event keys and dates; these are synthetic grounded tasks, not legal advice.
- Question and case splits
- Preserve 3,858/835/842 train/validation/test question rows and their underlying 1,106/237/238 case pairs. Do not randomly move related case evidence into another split.
- Evaluation information boundary
- Answers and label-available dates are evaluation metadata. Give only the evidence permitted by the task’s anchor; full retrospective timelines are not automatically valid prospective inputs.
- Custom benchmark scope
- A custom benchmark defines requested tasks, evidence windows, answer formats and grading rules in the accepted order; the reference task counts do not promise those order contents.
Methodology & Quality
- Data generation
- FCTE synthetic city engine
- Data classification
- Synthetic
FCTE generation methods and published city/domain checks provide source context. The order defines the final extract and the release-specific checks, documentation and limitations supplied with it.
Reference validation
These results describe the frozen Nairobi reference release SD-LAW-APPEALS-V1. Validation for a prepared custom extract is defined with that order.
The supplied Nairobi reference package; local integrity checks do not establish current public-repository byte equality or a future order’s certification
Publisher release gates
13/13 PASS
Supplied publisher validation report; not a claim that every documentation issue is resolved
Independent data rules
23/23 PASS
Reference joins, split membership, label consistency, temporal cutoffs and evidence resolution
Manifest entries
22/22 verified
Original local package bytes and SHA-256 hashes
Known schema issue
1 field / 2 tables
SequenceWithinDate: declared date, observed integer in full and pre-outcome timelines
Reference release modeling notes
- Grounded reference tasks
- The 5,535 JSONL questions cover appeal disposition, source-resolution effect/change and stay decision/scope. Answers resolve to supplied event keys and dates; these are synthetic grounded tasks, not legal advice.
- Question and case splits
- Preserve 3,858/835/842 train/validation/test question rows and their underlying 1,106/237/238 case pairs. Do not randomly move related case evidence into another split.
- Evaluation information boundary
- Answers and label-available dates are evaluation metadata. Give only the evidence permitted by the task’s anchor; full retrospective timelines are not automatically valid prospective inputs.
- Custom benchmark scope
- A custom benchmark defines requested tasks, evidence windows, answer formats and grading rules in the accepted order; the reference task counts do not promise those order contents.
Reference boundaries
- Synthetic cases are not observed legal cases, legal advice or representative outcome rates.
- Full case summary and case timeline are outcome-complete retrospective material; exclude final labels, later outcomes and retrospective context from prospective predictors.
- SequenceWithinDate is an observed integer ordering index (1–9), although the original dictionary declares date; derived stay rows can be blank. The supplied source dictionary has not been altered.
- Pre-outcome projection removes 12 source columns and adds three time-known stay replacements, producing 79 columns; do not treat the full 88-column timeline as leakage-safe.
- Absent ClosedDate or focal-citizen fields do not by themselves mean unresolved cases or invalid rows; this release documents case/party semantics and ambiguity.
- The supplied package contained no original public-use licence grant. The publisher selected NC public rights on 2026-10-13; current remote repository licence and byte parity were not verified.
- One Nairobi reference package supports these two primary profile associations; it does not certify the seven other related legal products or a custom paid extract.
- This reference supplies structured synthetic case events and grounded questions; it is not a corpus of court judgments, contracts or legal documents.
Known limitations
- Synthetic records describe generated scenarios; they do not establish observed population prevalence or real-world causal effects.
Delivery & License
- Catalogue base price
- $995 USD
- Price basis
- Base price per selected city
- Availability
- Prepared after order
The displayed catalogue amount is a base price per selected city. Your requested dates, population, fields, formats and usage rights determine the agreed scope and final quote. Applicable published discounts and taxes are confirmed with that quote.
Your prepared package
Your order confirms the data files, documentation, evidence, examples and release controls included in the prepared package.
- Agreed data extract
- Schema and field documentation
- Release-specific evidence and limitations
- Agreed loading or analysis examples
- Licence, manifest, checksums and delivery note
Package layout
Illustrative filenames. Your agreed order lists the final files and formats. Certification, benchmark splits and model examples are included only when specified.
SD_RESEARCH_DATASET/ ├── data/ │ ├── legal_records.parquet │ ├── contracts_records.parquet │ ├── construction_records.parquet │ └── business_records.parquet ├── tasks/ │ ├── task_definition.yaml │ ├── questions.jsonl │ ├── reference_answers.jsonl │ └── evidence_paths.jsonl ├── evaluation/ │ ├── baseline_model.py │ ├── metrics.json │ └── example_notebook.ipynb ├── documentation/ │ ├── README.md │ ├── DATA_DICTIONARY.csv │ ├── SCHEMA.md │ └── METHODOLOGY.md ├── evidence/ │ ├── CERTIFICATION.json │ ├── VALIDATION_REPORT.pdf │ └── KNOWN_LIMITATIONS.md ├── MANIFEST.json ├── CHECKSUMS.sha256 └── LICENSE.pdf
Delivery process
Confirm the cities, date range, records, formats, rights and price in writing. The order-specific extract is then prepared and supplied under the agreed payment and delivery conditions.
One secure download package for the agreed release and licensed entity.
Delivery timing, access expiry, formats and prerequisites are confirmed in the accepted order.
Support, refreshes, future releases, hosting and APIs are included only when expressly ordered.
Data Licensing Agreement · 2026-10
A non-exclusive, non-transferable licence for the named customer within the accepted scope and duration.
- Permitted use
- Internal research, development, testing and analytics within the accepted order.
- Model training
- Model training, external applications, commercial distribution and access by other legal entities must be expressly addressed in the order.
- Redistribution
- Raw or substantially reconstructable data cannot be resold, published or redistributed without separate written authorization.
- Attribution
- Preserve release provenance and applicable third-party notices. Published outputs must identify the data as synthetic.
Reference release license
The 2026-10-02 Nairobi reference uses CC-BY-NC-4.0. Existing rights granted for that named release remain available. Custom orders use the licensing terms confirmed for the agreed scope.
New publisher CC-BY-NC-4.0 choice for these publicly supplied reference materials. Original LICENSE_NOTICE.md contained no grant. Current remote repository licence and byte parity are not verified. Custom orders use separately agreed commercial terms.
New website dataset materials follow the publisher’s CC-BY-NC-4.0 policy. Previously released material retains its original license.
Reference citation file: CITATION.cff.
FAQ
Is this real personal data?
No. The records are synthetic and do not represent actual identifiable people or reproduce an individual's historical experience.
What does the catalogue price cover?
The displayed amount is a base price per selected city. The final scope and quote are agreed for your requested dates, population, fields, formats and rights. Adding a dataset to the cart starts the existing checkout inquiry; it does not purchase an unrestricted custom extract.
What can I inspect before ordering?
Download the verified reference sample and inspect the reference schema, checks, provenance and limitations under Data & Schema and Methodology & Quality. Its records and counts describe that named reference release; the illustrative custom-format preview is separate.
Is the dataset prepared for my requested dates?
Yes. Regular and Research datasets are prepared after an order, with the city, date range, population and other required scope agreed before preparation.
Can records be joined across domains?
The profile explains the intended relationships. The order-specific schema documents the supported keys, effective dates and joins for your agreed extract.
Can I use the dataset for AI or machine learning?
Review the study-specific inputs, outcomes and information cutoffs. The supplied Nairobi reference uses CC-BY-NC-4.0 under its stated reference licence scope; the current remote repository version has not been matched to these supplied bytes. Custom model training and external applications must be expressly addressed in the accepted order.
Can I use the data commercially?
Commercial use follows the accepted dataset-specific scope and licence. A sample download, cart selection, inquiry or consent checkbox does not activate a commercial licence or complete payment.
Can I request another city, period or format?
Yes. Use Data Request to describe the required city, population, dates, record concepts and formats. Onboarded city options for this product: Nairobi and Lagos. Other African major cities can be requested; availability, scope and the quote must be confirmed.
What documentation and evidence are included?
The accepted order confirms the schema, methodology, evidence, examples and release controls supplied with your prepared extract.
How do I receive the dataset?
After scope, terms and payment conditions are agreed, Fidelra Africa prepares the order-specific extract and supplies it through a secure download package. Timing and download-access expiry are defined in the order.
Does the public reference licence cover my custom order?
The supplied Nairobi reference uses CC-BY-NC-4.0 under the reference licence scope stated in this profile. A separately prepared custom extract follows the accepted order and the Data Licensing Agreement; reference counts and date coverage do not define that order.
Need different data?
Tell us the entities, fields, cities and time span you need.