CRILM (Contextually Relevant Imputation leveraging Language Models) transforms how we handle missing data in tabular datasets, replacing traditional numerical estimates with contextually relevant descriptors.

Key Innovations:

• Aligns tabular data with language models' natural strengths
• Uses large LMs to generate contextual descriptors for missing values
• Enables small LMs to effectively utilize these descriptors
• Bridges the gap between structured data analysis and natural language processing

Engineering Impact: This approach offers a paradigm shift for data preprocessing in engineering applications, potentially improving downstream analysis quality by preserving contextual relationships rather than forcing numerical approximations.

A Context-Aware Approach for Enhancing Data Imputation with Pre-trained Language Models