Web Tables for Few-Shot Language Model Fine-Tuning

The UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models. It was created by MicPie and last updated in August 2022.

Use Cases

• Fine-tune language models for Table Question Answering tasks using the provided web table examples.
• Improve model performance on Multi-Class Classification tasks structured from tabular data.
• Train models on Extractive QA tasks derived from the content of web tables.
• Benchmark few-shot learning capabilities on tasks like Natural Language Inference using tabular contexts.

Strengths

• Covers multiple task categories including Text Classification, Question Answering, and Language Modeling.
• Formatted specifically for few-shot learning, a distinct training paradigm.
• Last updated on 2022-08-04, providing a specific snapshot.

Limitations

• The specific number of rows, columns, and total data size is unknown.
• The raw table data and its original source are not described.
• The dataset's age may limit its relevance for evaluating on recent model architectures or benchmarks.

Provenance

Source

huggingface

Collection Method

Web tables formatted as few-shot tasks.

Freshness

Last updated 2022-08-04.