TAPEX: Table Pre-training via Learning a Neural SQL Executor
By Qian Liu, Jiaqi Guo, Zeqi Lin, Jian-Guang Lou. 2021
368 citations recorded by the source corpus.
Recent progress in language model pre-training has achieved a great success via leveraging large-scale unstructured textual data. However, it is still a challenge to apply pre-training on structured tabular data due to the absence of large-scale high-quality tabular data. In this paper, we propose TAPEX to show that table pre-training can be achieved by learning a neural SQL executor over a synthetic corpus, which is obtained by automatically synthesizing executable SQL queries and their execution outputs. TAPEX addresses the data scarcity challenge via guiding the language model to mimic a SQL executor on the diverse, large-scale and high-quality synthetic corpus. We evaluate TAPEX on four benchmark datasets. Experimental results demonstrate that TAPEX outperforms previous table pre-training approaches by a large margin and achieves new state-of-the-art results on all of them. This includes the improvements on the weakly-supervised WikiSQL denotation accuracy to 89.5% (+2.3%), the WikiTableQuestions denotation accuracy to 57.5% (+4.8%), the SQA denotation accuracy to 74.5% (+3.5%), and the TabFact accuracy to 84.2% (+3.2%). To our knowledge, this is the first work to exploit table pre-training via synthetic executable programs and to achieve new state-of-the-art results on various downstream tasks. Our code can be found at https://github.com/microsoft/Table-Pretraining.
Subfields: Artificial Intelligence, Information Systems, Retrieval-Augmented Generation, Natural Language Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Jianfeng Gao
- Luke S. Zettlemoyer
- Percy Liang
- Caiming Xiong
- Mohit Bansal
- Graham Neubig
- Richard Socher
- Noah A. Smith
In Information Systems
- Wenhu Chen
- Wen-tau Yih
- Dragomir R. Radev
- Xi Victoria Lin
- Tao Yu
- Caiming Xiong
- Jianfeng Gao
- Ji-rong Wen
In Retrieval-Augmented Generation
- Wenhu Chen
- Jianfeng Gao
- Michel Galley
- Luke S. Zettlemoyer
- Weizhu Chen
- Wen-tau Yih
- Noah A. Smith
- Ji-rong Wen
In Natural Language Processing
- Jianfeng Gao
- Caiming Xiong
- Luke S. Zettlemoyer
- Graham Neubig
- Percy Liang
- Richard Socher
- Wen-tau Yih
- Mohit Bansal
Citations in the corpus
- Reasoning Like Program Executors (2022, methodology): “Notably, following Liu et al. (2022), each database is flattened into a sequence when it is fed into LMs.”
- UnifiedSKG: Unifying and Multi-Tasking Structured Knowledge Grounding with Text-to-Text Language Models (2022, methodology): “Figure 2 illustrates how we convert the input x to an input sequence ˜ x and the output y to an output sequence ˜ y by means of “linearization” (Liu et al., 2021), enabling the unification of diverse forms of structured knowledge.”
- StructGPT: A General Framework for Large Language Model to Reason over Structured Data (2023, methodology): “For TableQA, we select MAPO (Liang et al., 2018), TAPAS (Herzig et al., 2020; Eisensch-los et al., 2020), UnifiedSKG (T5-3B) (Xie et al., 2022), TAPEX (Liu et al., 2022), and DATER (Ye et al., 2023).”
- Chameleon: Plug-and-Play Compositional Reasoning with Large Language Models (2023, background): “[28] Qian Liu, Bei Chen, Jiaqi Guo, Morteza Ziyadi, Zeqi Lin, Weizhu Chen, and Jian-Guang Lou.”
- LEVER: Learning to Verify Language-to-Code Generation with Execution (2023, methodology): “Following previous work (Xie et al., 2022; Liu et al., 2021; Ni et al., 2022; Zhang et al., 2022), we use execution accuracy as the main evaluation metric for all datasets, which measures the percentage of examples that yields the gold execution result or pass all test cases.”
- Binding Language Models in Symbolic Languages (2022, methodology): “4 Tapex (Liu et al., 2021) 85.9 Without Finetuning Codex end-to-end QA 72 .”
- A Survey of Deep Learning for Mathematical Reasoning (2022, neutral): “TaPEx (Liu et al., 2022b) pre-train their model based on BARTlarge, which has 460M parameters.”
- LIFT: Language-Interfaced Fine-Tuning for Non-Language Machine Learning Tasks (2022, background): “Several works also extend the existing LMs to handle different input data types, such as images [75, 76], audio [77], tabular data [78], and knowledge base [79] by updating the pretraining phase with these data and their corresponding tasks or using general-purpose architecture [80].”
- Dynamic Prompt Learning via Policy Gradient for Semi-structured Mathematical Reasoning (2022, positive): “TAPEX shows state-of-the-art performance on four table-related datasets.”
- Large Language Models are few(1)-shot Table Reasoners (2022, methodology): “In this paper, we mainly consider TAPAS (Herzig et al., 2020), TABERT (Yin et al., 2020), and TAPEX (Liu et al., 2021).”
- Griffin: Towards a Graph-Centric Relational Database Foundation Model (2025, unknown): “Tapex (Liu et al., 2021) explores learning a neural SQL executor.”