Taide Bench: 2,500 Samples for Five Language Model Evaluation Tasks

2,500 text samples across five tasks for evaluating language models, created by the author 'taide'. The dataset was last updated on 2024-04-12. It contains 500 samples each for letter writing, article writing, summarization, Chinese-to-English translation, and English-to-Chinese translation.

Use Cases

• Benchmarking text generation models based on the letter and article writing tasks
• Evaluating summarization model performance based on the summarization task
• Testing machine translation quality between Chinese and English based on the two translation tasks
• Assessing language model capabilities across multiple distinct text-based tasks

Strengths

• Contains 2,500 total samples across five distinct tasks
• Provides a balanced set of 500 samples per task for consistent evaluation
• Covers both Chinese and English languages for bilingual assessment

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Data may reflect bias inherent to the source platform and author's curation

Provenance

Source

huggingface

Freshness

Last updated 2024-04-12 00:43:43