Bangla LLM Evaluator Dataset
This dataset contains multiple evaluation splits in Bangla, each corresponding to a different evaluation domain (e.g., guardrails, toxicity, helpfulness). Each split is stored as a separate subset with the following columns:
- original_query: The original query in Bangla.
- translated_query: The query translated to English.
- generated_response: The response generated by the model (in English).
- translated_response: The response translated back to Bangla.
Each split is stored in its own subdirectory with a single Parquet file (test-00000-of-00001.parquet).
- Downloads last month
- 4