TAT-LLM: A Specialized Language Model for Discrete Reasoning over Tabular and Textual Data
This addresses the problem of efficient and secure discrete reasoning for applications like financial document analysis, though it is incremental as it builds on existing LLM fine-tuning methods.
The paper tackles question answering over hybrid tabular and textual data by developing TAT-LLM, a specialized language model that outperforms baseline models, including GPT-4, on benchmarks like FinQA, TAT-QA, and TAT-DQA.
In this work, we address question answering (QA) over a hybrid of tabular and textual data that are very common content on the Web (e.g. SEC filings), where discrete reasoning capabilities are often required. Recently, large language models (LLMs) like GPT-4 have demonstrated strong multi-step reasoning capabilities. We then consider harnessing the amazing power of LLMs to solve our task. We abstract a Step-wise Pipeline for tabular and textual QA, which consists of three key steps, including Extractor, Reasoner and Executor, and initially design an instruction to instantiate the pipeline and validate that GPT-4 outperforms all existing methods. However, utilizing an online LLM like GPT-4 holds various challenges in terms of cost, latency, and data security risk, which motivates us to specialize smaller LLMs in this task. We develop a TAT-LLM language model by fine-tuning LLaMA 2 with the training data generated automatically from existing expert-annotated datasets following the Step-wise Pipeline. The experimental results have verified that our TAT-LLM model can outperform all baseline models, including the previous best fine-tuned models and very large-scale LLMs like GPT-4 on FinQA, TAT-QA and TAT-DQA benchmarks.