Recent advances in natural language processing have leveraged instruction tuning to enhance Large Language Models (LLMs) for table-related tasks. However, previous works train different base models with different training data, lacking an apples-to-apples comparison across the result table LLMs. To address this, we fine-tune base models from the Mistral, OLMo, and Phi families on existing public training datasets. Our replication achieves performance on par with or surpassing existing table LLMs, establishing new state-of-the-art performance on Hitab, a table question-answering dataset. More importantly, through systematic out-of-domain evaluation, we decouple the contributions of training data and the base model, providing insight into their individual impacts. In addition, we assess the effects of table-specific instruction tuning on general-purpose benchmarks, revealing trade-offs between specialization and generalization.
@article{arxiv.2501.14717,
title = {Towards Better Understanding Table Instruction Tuning: Decoupling the Effects from Data versus Models},
author = {Naihao Deng and Sheng Zhang and Henghui Zhu and Shuaichen Chang and Jiani Zhang and Alexander Hanbo Li and Chung-Wei Hang and Hideo Kobayashi and Yiqun Hu and Patrick Ng},
journal= {arXiv preprint arXiv:2501.14717},
year = {2025}
}