AGIEval LogiQA-ZH: Chinese Logical Reasoning Questions for AI Benchmarking

AGIEval is a human-centric benchmark for evaluating foundation models. This dataset contains the Chinese-language LogiQA subtask, processed from the Microsoft AGIEval GitHub repository. It was authored by 'hails' and last updated on January 26, 2024.

Use Cases

• Benchmarking AI model performance on logical reasoning tasks based on the described Chinese LogiQA subtask.
• Training or fine-tuning models for Chinese question answering based on the logical problem structure.
• Studying cross-lingual transfer of reasoning skills between languages based on the parallel Chinese-English benchmark context.
• Analyzing the difficulty and composition of logic puzzles for AI evaluation based on the AGIEval framework.

Strengths

• Part of the established AGIEval benchmark from Microsoft, suggesting a standardized evaluation framework.
• Focuses specifically on Chinese-language logical reasoning, a noted capability gap for many AI models.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last updated 2024-01-26 18:41:16; freshness should be verified.

Provenance

Source

Microsoft AGIEval GitHub repository (https://github.com/microsoft/AGIEval).

Collection Method

Processed as in the original repository, following the pattern of other AGIEval datasets on Hugging Face.

Freshness

Last updated 2024-01-26 18:41:16.