dnaihao/phi-3-mini-tablellm
dnaihao/phi-3-mini-tablellm is a 4 billion parameter language model, a replication of TableLLM, fine-tuned from Microsoft's Phi-3-mini-4k-instruct. It is specifically trained on a table-centric instruction-tuning corpus to enhance its capabilities in understanding and processing tabular data. This model is designed for tasks involving structured information extraction and manipulation from tables, serving as an artifact for research into model and data effects on Table LLMs.
Loading preview...
Overview
dnaihao/phi-3-mini-tablellm is a 4 billion parameter language model, replicating the TableLLM architecture. It is fine-tuned from Microsoft's Phi-3-mini-4k-instruct base model, utilizing a specialized instruction-tuning corpus focused on tabular data. This model was released as an additional artifact alongside the EACL 2026 Findings paper "What Really Matters for Table LLMs? A Meta-Evaluation of Model and Data Effects" by Deng et al., extending the paper's experimental grid.
Key Capabilities
- Table-specific understanding: Enhanced ability to process and interpret structured information within tables.
- Instruction-following for tabular tasks: Fine-tuned on
tablellm_train.jsonfrom thednaihao/Table-Instructsdataset, making it proficient in responding to instructions related to tables. - Research artifact: Provides a specific base-model variant for meta-evaluation of model and data effects on Table LLMs, complementing broader research on 7B-scale models.
Good for
- Research in Table LLMs: Ideal for researchers studying the impact of different base models and training data on table-related tasks.
- Structured data extraction: Applications requiring the extraction of specific information from tables based on natural language instructions.
- Benchmarking table-centric tasks: Useful for evaluating model performance on datasets involving tabular data processing.