Persian MMLU & Math Benchmark: GPT-OSS-120B Evaluation Results

Persian-language evaluation results for the openai/gpt-oss-120b model on a suite of benchmarks. The dataset covers general knowledge, mathematical reasoning, and language understanding tasks, using the first 15 samples per sub-task with 0-shot prompting. It was created by author artindnr and last updated on Hugging Face in July 2026.

Use Cases

• Benchmarking model performance on Persian general knowledge based on the Persian MMLU tasks mentioned in the description
• Evaluating mathematical reasoning capabilities in Persian based on the Persian Math benchmark
• Assessing language understanding on specific tasks like entailment, machine translation, and reading comprehension based on the included ParsiNLU tasks

Strengths

• Focuses on Persian-language AI evaluation, a less common language in benchmarks
• Evaluates a specific large model (openai/gpt-oss-120b) with a defined prompting strategy (0-shot)
• Covers multiple task types: general knowledge, mathematical reasoning, and language understanding

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download
• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

huggingface

Collection Method

Likely contains evaluation results from running the specified model on the mentioned benchmarks.

Freshness

Last updated 2026-07-22 00:52:52; freshness should be verified