JOR-Bench: Japanese Operations Research Benchmarks for LLM Evaluation

JOR-Bench is a bilingual evaluation benchmark containing 1,319 Operations Research word problems in English and Japanese. It was created by translating problems from five publicly available English benchmarks. The dataset was authored by cyberagent and last updated on July 8, 2026.

Use Cases

• Benchmarking LLM performance on translating natural language to mathematical formulations based on the described word problems.
• Evaluating cross-lingual reasoning capabilities of LLMs using the bilingual English-Japanese problem pairs.
• Training or fine-tuning models for Operations Research problem-solving based on the provided problem set.
• Studying the difficulty and structure of OR problems for AI systems based on the benchmark's composition.

Strengths

• Contains 1,319 distinct word problems for evaluation.
• Provides bilingual (English and Japanese) versions of each problem.
• Derived from five publicly available source benchmarks, suggesting a degree of curation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Freshness should be verified as the last update timestamp is in the future (2026-07-08).

Provenance

Source

Translated from five publicly available English benchmarks.

Collection Method

Derived via Japanese translation of existing English benchmarks.

Freshness

Last updated 2026-07-08 11:46:41