Cultural Knowledge Benchmarks
Original questions on customs, etiquette and regional knowledge, written to avoid benchmark contamination.
Pairwise-judgment schematic. Illustration only. This is not sample data, and no sample exists until one is produced for your request.
- Ready to License
- Partner Sourced
- Custom Collection
Recorded or built to your specification, per project.
- Sourcing on Request
Specification, volume and rights for this entry are not yet fixed and are defined with you.
- Data formats
- JSONL
- Annotation
- Answer key, Topic
- Language
- Japanese (ja)
- Region
- Japan (scope to be defined per project)
- Modality
- Text
- Volume
- Not specified. Set per project.
- Price
- Quoted per project.
Target models
LLM evaluation
Use cases
Model Evaluation · Large Language Models
How it is collected
Collected or authored to your specification by recruited Japanese participants or experts, under written consent and a project-specific protocol.
Quality assurance
Native-speaker review of a sample set before delivery, schema and format validation, duplicate checks, and a written QA report agreed in advance.
Licensing & usage
Delivered under a written license or work-for-hire agreement defining permitted AI training and evaluation use, territory, term and any exclusivity. Contributor consent is documented.
Need this, or something adjacent?
Describe the data you need. We will respond with feasibility, rights approach and next steps.