pith. sign in

arxiv: 2502.12911 · v3 · submitted 2025-02-18 · 💻 cs.CL · cs.DB

Knapsack Optimization-based Schema Linking for LLM-based Text-to-SQL Generation

Zheng Yuan , Hao Chen , Zijin Hong , Qinggang Zhang , Feiran Huang , Qing Li , Xiao Huang This is my paper

Pith reviewed 2026-05-23 02:39 UTC · model grok-4.3

classification 💻 cs.CL cs.DB
keywords schema linkingtext-to-SQLknapsack optimizationLLMSpider benchmarkBIRD benchmarkSQL generation
0
0 comments X

The pith

Framing schema linking as a knapsack problem lets a 1.6B model outperform larger LLMs on text-to-SQL tasks.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proposes KaSLA as a plug-in schema linking method that treats selection of relevant database tables and columns as a knapsack optimization problem. It adds a restricted missing indicator to standard metrics so that missing key elements is penalized even when precision stays high. A hierarchical process first solves the knapsack for tables then for columns inside selected tables, allowing limited tolerance for redundant items. This setup produces better linking results than large language models and raises end-to-end SQL generation accuracy when substituted into existing Text2SQL pipelines on the Spider and BIRD benchmarks.

Core claim

KaSLA employs knapsack optimization within a hierarchical linking process to identify optimal tables and then columns, incorporating a tolerance for potentially redundant elements to prevent missing relevant schema items. This results in superior schema linking performance compared to state-of-the-art methods used with larger models. Substituting KaSLA into SOTA Text2SQL models yields measurable gains on Spider and BIRD.

What carries the argument

Knapsack optimization with hierarchical table-then-column selection and a restricted missing indicator that penalizes omissions of relevant schema elements

If this is right

  • KaSLA can replace the schema linking stage inside existing Text2SQL systems and raise their SQL generation accuracy.
  • A 1.6B-parameter model using KaSLA produces better schema links than DeepSeek-V3 paired with prior SOTA linking methods.
  • The hierarchical knapsack approach reduces both missing relevant elements and excess redundant ones.
  • Performance gains hold on the Spider and BIRD benchmarks when KaSLA is substituted into SOTA models.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same knapsack framing could be tested on schema linking for other query languages or API retrieval tasks where omission of key fields is costly.
  • If relevance scores fed into the knapsack are improved, the tolerance parameter might become less sensitive across domains.
  • KaSLA's small model size suggests it could lower overall latency in production Text2SQL pipelines that currently rely on large models for every stage.

Load-bearing premise

Modeling schema linking as a knapsack problem with a tunable tolerance for redundant elements will reliably avoid missing relevant schema items across unseen databases and query distributions without benchmark-specific retuning.

What would settle it

Apply the tolerance value tuned on Spider to an entirely new database schema and query distribution, then measure whether the rate of missed relevant elements rises above the rate observed on Spider.

Figures

Figures reproduced from arXiv: 2502.12911 by Feiran Huang, Hao Chen, Qinggang Zhang, Qing Li, Xiao Huang, Zheng Yuan, Zijin Hong.

Figure 1
Figure 1. Figure 1: Performance comparison of different schema linking methods on the [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Commonly used metrics, like Recall and Precision, fail to accurately [PITH_FULL_IMAGE:figures/full_fig_p001_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The imbalanced proportion of relevant schema elements with the [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Missing & redundancy seesaw problem in schema linking. We evaluate [PITH_FULL_IMAGE:figures/full_fig_p003_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Overall framework of proposed knapsack optimization-based schema linking approach (KaSLA) and and how it operates within a text-to-SQL system. [PITH_FULL_IMAGE:figures/full_fig_p004_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: (a) Execution accuracy (EX, %) of CodeS + KaSLA on BIRD-dev [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: The seen and unseen data in BIRD-dev relative to BIRD-train. [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Failure case analysis of “CodeS + KaSLA” in BRD-dev. [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
read the original abstract

Generating SQLs from user queries is a long-standing challenge, where the accuracy of initial schema linking significantly impacts subsequent SQL generation performance. However, current schema linking models still struggle with missing relevant schema elements or an excess of redundant ones. A crucial reason for this is that commonly used metrics, recall and precision, fail to capture relevant element missing and thus cannot reflect actual schema linking performance. Motivated by this, we propose enhanced schema linking metrics by introducing a \textbf{restricted missing indicator}. Accordingly, we introduce \textbf{\underline{K}n\underline{a}psack optimization-based \underline{S}chema \underline{L}inking \underline{A}pproach (KaSLA)}, a plug-in schema linking method designed to prevent the missing of relevant schema elements while minimizing the inclusion of redundant ones. KaSLA employs a hierarchical linking strategy that first identifies the optimal table linking and subsequently links columns within the selected table to reduce linking candidate space. In each linking process, it utilizes a knapsack optimization approach to link potentially relevant elements while accounting for a limited tolerance of potentially redundant ones. With this optimization, KaSLA-1.6B achieves superior schema linking results compared to large-scale LLMs, including DeepSeek-V3 with the state-of-the-art (SOTA) schema linking method. Extensive experiments on Spider and BIRD benchmarks verify that KaSLA can significantly improve the SQL generation performance of SOTA Text2SQL models by substituting their schema linking processes. The code is available at https://github.com/DEEP-PolyU/KaSLA.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 3 minor

Summary. The manuscript proposes KaSLA, a plug-in schema linking method for Text-to-SQL that models the task as a knapsack optimization problem. It introduces a restricted missing indicator to better evaluate linking quality beyond recall/precision, uses a hierarchical (table-then-column) strategy to reduce search space, and applies knapsack selection that tolerates limited redundancy while aiming to avoid missing relevant schema elements. The 1.6B-parameter KaSLA is claimed to outperform large LLMs including DeepSeek-V3 (with SOTA schema linking) on linking quality and to improve end-to-end SQL generation of SOTA Text2SQL models when substituted on Spider and BIRD benchmarks. Code is released.

Significance. If the central claims hold after addressing the robustness issues, the work is significant: schema linking remains a key bottleneck for Text-to-SQL, and a lightweight, optimization-based plug-in that beats much larger models while releasing code would be a practical contribution. The restricted missing indicator and hierarchical knapsack formulation offer a principled alternative to heuristic linking, with potential for broader use if the tolerance parameter generalizes.

major comments (3)
  1. [§3.2 and §4.1] §3.2 (restricted missing indicator definition) and §4.1 (knapsack formulation): the central claim that KaSLA reliably prevents relevant schema misses rests on the restricted missing indicator and the knapsack objective with its redundancy tolerance budget; however, the manuscript provides no sensitivity analysis or cross-distribution validation showing that a single tolerance value avoids misses on schemas outside the Spider/BIRD training distribution, which directly undermines the plug-in superiority claim over DeepSeek-V3.
  2. [§5.2] §5.2 (experimental results on BIRD): the reported gains when substituting KaSLA into SOTA Text2SQL models are load-bearing for the end-to-end claim, yet no ablation isolates whether performance depends on benchmark-specific choice of the redundancy tolerance (a free parameter) versus the knapsack structure itself; without this, the results cannot be interpreted as evidence of distribution-agnostic reliability.
  3. [Table 3] Table 3 (schema linking metrics): the comparison showing KaSLA-1.6B superior to DeepSeek-V3 is central, but the table does not report the exact tolerance value used for each method or whether it was held constant across Spider and BIRD; this leaves open whether the reported superiority requires per-benchmark adjustment.
minor comments (3)
  1. [Abstract] Abstract: the hierarchical linking strategy and knapsack objective are described at a high level, but the precise objective function (value vs. weight definitions) should be stated explicitly in the abstract for immediate clarity.
  2. [§2] §2 (related work): several recent schema linking papers using LLM prompting are cited, but the discussion does not contrast the knapsack formulation against other optimization-based linking methods that have appeared in the broader database literature.
  3. [Figure 2] Figure 2 (KaSLA pipeline diagram): the diagram would benefit from explicit annotation of the redundancy tolerance input and the restricted missing indicator computation to make the flow self-contained.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below and will revise the manuscript accordingly to strengthen the robustness claims.

read point-by-point responses

  1. Referee: [§3.2 and §4.1] §3.2 (restricted missing indicator definition) and §4.1 (knapsack formulation): the central claim that KaSLA reliably prevents relevant schema misses rests on the restricted missing indicator and the knapsack objective with its redundancy tolerance budget; however, the manuscript provides no sensitivity analysis or cross-distribution validation showing that a single tolerance value avoids misses on schemas outside the Spider/BIRD training distribution, which directly undermines the plug-in superiority claim over DeepSeek-V3.

    Authors: We agree that sensitivity analysis would strengthen the generalizability claim. The manuscript uses a single fixed tolerance value selected via validation on Spider dev and applied unchanged to BIRD. In the revision we will add a dedicated sensitivity analysis subsection evaluating performance across a range of tolerance values on both benchmarks plus an additional out-of-distribution schema set to demonstrate stability. revision: yes

  2. Referee: [§5.2] §5.2 (experimental results on BIRD): the reported gains when substituting KaSLA into SOTA Text2SQL models are load-bearing for the end-to-end claim, yet no ablation isolates whether performance depends on benchmark-specific choice of the redundancy tolerance (a free parameter) versus the knapsack structure itself; without this, the results cannot be interpreted as evidence of distribution-agnostic reliability.

    Authors: We will add an ablation in the revised §5.2 that compares the full hierarchical knapsack formulation against a non-optimization variant that applies the identical tolerance threshold, thereby isolating the contribution of the knapsack objective from the choice of tolerance value. revision: yes

  3. Referee: [Table 3] Table 3 (schema linking metrics): the comparison showing KaSLA-1.6B superior to DeepSeek-V3 is central, but the table does not report the exact tolerance value used for each method or whether it was held constant across Spider and BIRD; this leaves open whether the reported superiority requires per-benchmark adjustment.

    Authors: We will update Table 3 with a footnote stating the exact tolerance value employed for KaSLA and confirming it was held constant across both benchmarks; for DeepSeek-V3 we will reference the configuration reported in its original work. revision: yes

Circularity Check

0 steps flagged

No circularity: method is an independent algorithmic formulation evaluated on external benchmarks

full rationale

The paper presents KaSLA as a new knapsack-based optimization procedure for schema linking, introducing a restricted missing indicator metric and a hierarchical table-then-column strategy with tunable tolerance. No equations or claims reduce by construction to fitted parameters or self-citations; performance claims rest on direct evaluation against Spider and BIRD benchmarks using external SOTA baselines (including DeepSeek-V3). The derivation chain is self-contained as an applied optimization technique without load-bearing self-references or renaming of prior results.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claim rests on the modeling choice that schema linking can be cast as a knapsack problem and on the new restricted missing indicator; no external machine-checked proofs or parameter-free derivations are referenced.

free parameters (1)
  • redundancy tolerance budget
    The limited tolerance for potentially redundant elements is a tunable capacity parameter in the knapsack formulation that directly affects which schema items are selected.
axioms (1)
  • domain assumption Schema linking relevance and redundancy can be scored such that a knapsack solver yields near-optimal links without missing critical elements.
    This modeling premise is invoked when the paper states that the knapsack approach prevents missing relevant elements while minimizing redundant ones.

pith-pipeline@v0.9.0 · 5828 in / 1309 out tokens · 45091 ms · 2026-05-23T02:39:19.582137+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Both Ends Count! Just How Good are LLM Agents at "Text-to-Big SQL"?

    cs.DB 2026-02 unverdicted novelty 7.0

    New Text-to-Big SQL metrics show that LLM agents must balance accuracy with cost and speed at scale, where GPT-4o trades some accuracy for up to 12x speedup and GPT-5.2 proves more cost-effective than Gemini 3 Pro on ...

  2. EHRAG: Bridging Semantic Gaps in Lightweight GraphRAG via Hybrid Hypergraph Construction and Retrieval

    cs.AI 2026-04 unverdicted novelty 6.0

    EHRAG constructs structural hyperedges from sentence co-occurrence and semantic hyperedges from entity embedding clusters, then applies hybrid diffusion plus topic-aware PPR to retrieve top-k documents, outperforming ...

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages · cited by 2 Pith papers · 9 internal anchors

  1. [1]

    GPT-4 Technical Report

    J. Achiam, S. Adler, S. Agarwal, L. Ahmad, I. Akkaya, F. L. Aleman, D. Almeida, J. Altenschmidt, S. Altman, S. Anadkat et al., “Gpt-4 technical report,” arXiv preprint arXiv:2303.08774, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  2. [2]

    The Llama 3 Herd of Models

    A. Dubey, A. Jauhri, A. Pandey, A. Kadian, A. Al-Dahle, A. Letman, A. Mathur, A. Schelten, A. Yang, A. Fan et al. , “The llama 3 herd of models,” arXiv preprint arXiv:2407.21783, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  3. [3]

    Next-generation database interfaces: A survey of llm-based text-to-sql,

    Z. Hong, Z. Yuan, Q. Zhang, H. Chen, J. Dong, F. Huang, and X. Huang, “Next-generation database interfaces: A survey of llm-based text-to-sql,” arXiv preprint arXiv:2406.08426, 2024

    work page arXiv 2024

  4. [4]

    Text-to-sql empowered by large language mod- els: A benchmark evaluation,

    D. Gao, H. Wang, Y . Li, X. Sun, Y . Qian, B. Ding, and J. Zhou, “Text-to-sql empowered by large language mod- els: A benchmark evaluation,” in International Conference on Very Large Data Bases (VLDB) , 2024

    work page 2024

  5. [5]

    Before generation, align it! a novel and effective strategy for mitigating hallucinations in text-to-sql generation,

    G. Qu, J. Li, B. Li, B. Qin, N. Huo, C. Ma, and R. Cheng, “Before generation, align it! a novel and effective strategy for mitigating hallucinations in text-to-sql generation,” arXiv preprint arXiv:2405.15307 , 2024

    work page arXiv 2024

  6. [6]

    The death of schema linking? text-to-sql in the age of well-reasoned language models,

    K. Maamari, F. Abubaker, D. Jaroslawicz, and A. Mhed- hbi, “The death of schema linking? text-to-sql in the age of well-reasoned language models,” arXiv preprint arXiv:2408.07702, 2024

    work page arXiv 2024

  7. [7]

    Resdsql: Decoupling schema linking and skeleton parsing for text-to-sql,

    H. Li, J. Zhang, C. Li, and H. Chen, “Resdsql: Decoupling schema linking and skeleton parsing for text-to-sql,” in Conference on Artificial Intelligence (AAAI) , 2023

    work page 2023

  8. [8]

    Codes: Towards building open-source language models for text-to-sql,

    H. Li, J. Zhang, H. Liu, J. Fan, X. Zhang, J. Zhu, R. Wei, H. Pan, C. Li, and H. Chen, “Codes: Towards building open-source language models for text-to-sql,” in Conference on Management of Data (SIGMOD) , 2024

    work page 2024

  9. [9]

    Dts-sql: Decomposed text- to-sql with small large language models,

    M. Pourreza and D. Rafiei, “Dts-sql: Decomposed text- to-sql with small large language models,” arXiv preprint arXiv:2402.01117, 2024

    work page arXiv 2024

  10. [10]

    Solid-SQL: Enhanced schema-linking based in- context learning for robust text-to-SQL,

    G. Liu, Y . Tan, R. Zhong, Y . Xie, L. Zhao, Q. Wang, B. Hu, and Z. Li, “Solid-SQL: Enhanced schema-linking based in- context learning for robust text-to-SQL,” in Proceedings of the 31st International Conference on Computational Linguistics. Association for Computational Linguistics, 2025, pp. 9793–9803

    work page 2025

  11. [11]

    CHESS: Contextual Harnessing for Efficient SQL Synthesis

    S. Talaei, M. Pourreza, Y .-C. Chang, A. Mirhoseini, and A. Saberi, “Chess: Contextual harnessing for efficient sql synthesis,” arXiv preprint arXiv:2405.16755 , 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  12. [12]

    Rsl-sql: Robust schema linking in text-to-sql generation,

    Z. Cao, Y . Zheng, Z. Fan, X. Zhang, and W. Chen, “Rsl-sql: Robust schema linking in text-to-sql generation,” arXiv preprint arXiv:2411.00073 , 2024

    work page arXiv 2024

  13. [13]

    The multidimensional 0–1 knapsack problem: An overview,

    A. Fr´eville, “The multidimensional 0–1 knapsack problem: An overview,”European Journal of Operational Research, 2004

    work page 2004

  14. [14]

    DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligence

    D. Guo, Q. Zhu, D. Yang, Z. Xie, K. Dong, W. Zhang, G. Chen, X. Bi, Y . Wu, Y . Li et al. , “Deepseek-coder: When the large language model meets programming–the rise of code intelligence,”arXiv preprint arXiv:2401.14196, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  15. [15]

    LoRA: Low-Rank Adaptation of Large Language Models

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, and W. Chen, “Lora: Low-rank adaptation of large language models,” arXiv preprint arXiv:2106.09685, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  16. [16]

    RoBERTa: A Robustly Optimized BERT Pretraining Approach

    Y . Liu, M. Ott, N. Goyal, J. Du, M. Joshi, D. Chen, O. Levy, M. Lewis, L. Zettlemoyer, and V . Stoyanov, “Roberta: A robustly optimized bert pretraining approach,” arXiv preprint arXiv:1907.11692 , 2019

    work page internal anchor Pith review Pith/arXiv arXiv 1907

  17. [17]

    Focal loss for dense object detection,

    T.-Y . Ross and G. Doll´ar, “Focal loss for dense object detection,” in proceedings of the IEEE conference on computer vision and pattern recognition , 2017, pp. 2980– 2988

    work page 2017

  18. [18]

    SimCSE: Simple Contrastive Learning of Sentence Embeddings

    T. Gao, X. Yao, and D. Chen, “Simcse: Simple con- trastive learning of sentence embeddings,” arXiv preprint arXiv:2104.08821, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  19. [19]

    Can LLM already serve as a database interface? a BIg bench for large-scale database grounded text-to-SQLs,

    J. Li, B. Hui, G. QU, J. Yang, B. Li, B. Li, B. Wang, B. Qin, R. Geng, N. Huo, X. Zhou, C. Ma, G. Li, K. Chang, F. Huang, R. Cheng, and Y . Li, “Can LLM already serve as a database interface? a BIg bench for large-scale database grounded text-to-SQLs,” in Advances in Neural Information Processing Systems (NeurIPS) , 2023

    work page 2023

  20. [20]

    Spider: A large-scale human-labeled dataset for complex and cross-domain semantic parsing and text-to-SQL task,

    T. Yu, R. Zhang, K. Yang, M. Yasunaga, D. Wang, Z. Li, J. Ma, I. Li, Q. Yao, S. Roman, Z. Zhang, and D. Radev, “Spider: A large-scale human-labeled dataset for complex and cross-domain semantic parsing and text-to-SQL task,” in Empirical Methods in Natural Language Processing (EMNLP), 2018

    work page 2018

  21. [21]

    Dawei Gao, Haibin Wang, Yaliang Li, Xiuyu Sun, Yichen Qian, Bolin Ding, and Jingren Zhou

    X. Dong, C. Zhang, Y . Ge, Y . Mao, Y . Gao, J. Lin, D. Lou et al., “C3: Zero-shot text-to-sql with chatgpt,” arXiv preprint arXiv:2307.07306 , 2023

    work page arXiv 2023

  22. [22]

    DIN-SQL: Decomposed in-context learning of text-to-SQL with self-correction,

    M. Pourreza and D. Rafiei, “DIN-SQL: Decomposed in-context learning of text-to-SQL with self-correction,” in Advances in Neural Information Processing Systems (NeurIPS), 2023

    work page 2023

  23. [23]

    Mac- sql: A multi-agent collaborative framework for text-to-sql,

    B. Wang, C. Ren, J. Yang, X. Liang, J. Bai, L. Chai, Z. Yan, Q.-W. Zhang, D. Yin, X. Sun, and Z. Li, “Mac- sql: A multi-agent collaborative framework for text-to-sql,” arXiv preprint arXiv:2312.11242 , 2024

    work page arXiv 2024

  24. [24]

    The dawn of natural language to sql: Are we fully ready?

    B. Li, Y . Luo, C. Chai, G. Li, and N. Tang, “The dawn of natural language to sql: Are we fully ready?” arXiv preprint arXiv:2406.01265, 2024

    work page arXiv 2024

  25. [25]

    Dubo- sql: Diverse retrieval-augmented generation and fine tuning for text-to-sql,

    D. G. Thorpe, A. J. Duberstein, and I. A. Kinsey, “Dubo- sql: Diverse retrieval-augmented generation and fine tuning for text-to-sql,” arXiv preprint arXiv:2404.12560 , 2024

    work page arXiv 2024

  26. [26]

    E-sql: Direct schema linking via question enrichment in text-to-sql,

    H. A. Cafero ˘glu and ¨O. Ulusoy, “E-sql: Direct schema linking via question enrichment in text-to-sql,” arXiv preprint arXiv:2409.16751, 2024

    work page arXiv 2024

  27. [27]

    DeepSeek-V3 Technical Report

    A. Liu, B. Feng, B. Xue, B. Wang, B. Wu, C. Lu, C. Zhao, C. Deng, C. Zhang, C. Ruan et al., “Deepseek-v3 technical report,” arXiv preprint arXiv:2412.19437 , 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  28. [28]

    StarCoder 2 and The Stack v2: The Next Generation

    A. Lozhkov, R. Li, L. B. Allal, F. Cassano, J. Lamy- Poirier, N. Tazi, A. Tang, D. Pykhtar, J. Liu, Y . Weiet al., “Starcoder 2 and the stack v2: The next generation,” arXiv preprint arXiv:2402.19173, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  29. [29]

    Prompt programming for large language models: Beyond the few-shot paradigm,

    L. Reynolds and K. McDonell, “Prompt programming for large language models: Beyond the few-shot paradigm,” in Conference on Human Factors in Computing Systems (CHI), 2021

    work page 2021

  30. [30]

    A comprehensive evaluation of chatgpt’s zero-shot text-to-sql capability,

    A. Liu, X. Hu, L. Wen, and P. S. Yu, “A comprehensive evaluation of chatgpt’s zero-shot text-to-sql capability,” arXiv preprint arXiv:2303.13547 , 2023

    work page arXiv 2023

  31. [31]

    Chain-of-thought prompting elicits reasoning in large language models,

    J. Wei, X. Wang, D. Schuurmans, M. Bosma, F. Xia, E. Chi, Q. V . Le, D. Zhou et al. , “Chain-of-thought prompting elicits reasoning in large language models,” in Advances in Neural Information Processing Systems (NeurIPS), 2022

    work page 2022

  32. [32]

    Finetuned language models are zero-shot learners,

    J. Wei, M. Bosma, V . Y . Zhao, K. Guu, A. W. Yu, B. Lester, N. Du, A. M. Dai, and Q. V . Le, “Finetuned language models are zero-shot learners,” in International Conference on Learning Representations (ICLR) , 2021

    work page 2021

  33. [33]

    Exploring chain of thought style prompting for text- to-SQL,

    C.-Y . Tai, Z. Chen, T. Zhang, X. Deng, and H. Sun, “Exploring chain of thought style prompting for text- to-SQL,” in Empirical Methods in Natural Language Processing (EMNLP), 2023

    work page 2023

  34. [34]

    ACT-SQL: In-context learning for text-to-SQL with automatically- generated chain-of-thought,

    H. Zhang, R. Cao, L. Chen, H. Xu, and K. Yu, “ACT-SQL: In-context learning for text-to-SQL with automatically- generated chain-of-thought,” in Findings of Empirical Methods in Natural Language Processing (EMNLP) , 2023

    work page 2023

  35. [35]

    Knowledge-to-sql: Enhancing sql generation with data expert llm,

    Z. Hong, Z. Yuan, H. Chen, Q. Zhang, F. Huang, and X. Huang, “Knowledge-to-sql: Enhancing sql generation with data expert llm,” in Findings of Association for Computational Linguistics (ACL) , 2024

    work page 2024

  36. [36]

    Chase-sql: Multi-path reasoning and preference optimized candidate selection in text-to-sql,

    M. Pourreza, H. Li, R. Sun, Y . Chung, S. Talaei, G. T. Kakkar, Y . Gan, A. Saberi, F. Ozcan, and S. O. Arik, “Chase-sql: Multi-path reasoning and preference optimized candidate selection in text-to-sql,” in International Con- ference on Learning Representations (ICLR) , 2024

    work page 2024

  37. [37]

    Towards robustness of text-to-SQL models against synonym substitution,

    Y . Gan, X. Chen, Q. Huang, M. Purver, J. R. Woodward, J. Xie, and P. Huang, “Towards robustness of text-to-SQL models against synonym substitution,” in Association for Computational Linguistics and International Joint Con- ference on Natural Language Processing (ACL-IJCNLP) , 2021

    work page 2021

  38. [38]

    Structure-grounded pretraining for text-to-SQL,

    X. Deng, A. H. Awadallah, C. Meek, O. Polozov, H. Sun, and M. Richardson, “Structure-grounded pretraining for text-to-SQL,” in North American Chapter of the Associ- ation for Computational Linguistics: Human Language Technologies (NAACL-HLT), 2021

    work page 2021

  39. [39]

    Purple: Making a large language model a better sql writer,

    T. Ren, Y . Fan, Z. He, R. Huang, J. Dai, C. Huang, Y . Jing, K. Zhang, Y . Yang, and X. S. Wang, “Purple: Making a large language model a better sql writer,” in International Conference on Data Engineering (ICDE) , 2024

    work page 2024

  40. [40]

    Xiyan-sql: A multi- generator ensemble framework for text-to-sql,

    Y . Gao, Y . Liu, X. Li, X. Shi, Y . Zhu, Y . Wang, S. Li, W. Li, Y . Hong, Z. Luo et al. , “Xiyan-sql: A multi- generator ensemble framework for text-to-sql,” arXiv preprint arXiv:2411.08599, 2024

    work page arXiv 2024

  41. [41]

    Structure guided large language model for sql generation,

    Q. Zhang, J. Dong, H. Chen, W. Li, F. Huang, and X. Huang, “Structure guided large language model for sql generation,” arXiv preprint arXiv:2402.13284 , 2024

    work page arXiv 2024

  42. [42]

    Benchmarking large language models via random variables,

    Z. Hong, H. Wu, S. Dong, J. Dong, Y . Xiao, Y . Zhang, Z. Wang, F. Huang, L. Li, H. Yang et al., “Benchmarking large language models via random variables,” arXiv preprint arXiv:2501.11790, 2025

    work page arXiv 2025

  43. [43]

    Macro graph neural networks for online billion-scale recommender systems,

    H. Chen, Y . Bei, Q. Shen, Y . Xu, S. Zhou, W. Huang, F. Huang, S. Wang, and X. Huang, “Macro graph neural networks for online billion-scale recommender systems,” in Proceedings of the ACM on Web Conference 2024 , 2024, pp. 3598–3608

    work page 2024

  44. [44]

    A survey of graph retrieval-augmented generation for customized large language models,

    Q. Zhang, S. Chen, Y . Bei, Z. Yuan, H. Zhou, Z. Hong, J. Dong, H. Chen, Y . Chang, and X. Huang, “A survey of graph retrieval-augmented generation for customized large language models,” arXiv preprint arXiv:2501.13958, 2025

    work page arXiv 2025

  45. [45]

    Large language model interaction simulator for cold-start item recommendation,

    F. Huang, Z. Yang, J. Jiang, Y . Bei, Y . Zhang, and H. Chen, “Large language model interaction simulator for cold-start item recommendation,” arXiv preprint arXiv:2402.09176 , 2024

    work page arXiv 2024

  46. [46]

    Proton: Probing schema linking information from pre-trained language models for text-to-sql parsing,

    L. Wang, B. Qin, B. Hui, B. Li, M. Yang, B. Wang, B. Li, J. Sun, F. Huang, L. Si, and Y . Li, “Proton: Probing schema linking information from pre-trained language models for text-to-sql parsing,” in Conference on Knowledge Discovery and Data Mining (KDD) , 2022

    work page 2022

  47. [47]

    A survey on text-to-sql parsing: Concepts, methods, and future directions,

    B. Qin, B. Hui, L. Wang, M. Yang, J. Li, B. Li, R. Geng, R. Cao, J. Sun, L. Si et al. , “A survey on text-to-sql parsing: Concepts, methods, and future directions,” arXiv preprint arXiv:2208.13629, 2022

    work page arXiv 2022

  48. [48]

    Constructing an interactive natural language interface for relational databases,

    F. Li and H. V . Jagadish, “Constructing an interactive natural language interface for relational databases,” in International Conference on Very Large Data Bases (VLDB), 2014

    work page 2014

  49. [49]

    A rule based approach for nlp based query processing,

    T. Mahmud, K. A. Hasan, M. Ahmed, and T. H. C. Chak, “A rule based approach for nlp based query processing,” in International Conference on Electrical Information and Communication Technologies (EICT), 2015

    work page 2015

  50. [50]

    Grappa: Grammar-augmented pre-training for table se- mantic parsing,

    T. Yu, C.-S. Wu, X. V . Lin, bailin wang, Y . C. Tan, X. Yang, D. Radev, richard socher, and C. Xiong, “Grappa: Grammar-augmented pre-training for table se- mantic parsing,” in International Conference on Learning Representations (ICLR), 2021

    work page 2021

  51. [51]

    A survey on deep learning approaches for text-to-sql,

    G. Katsogiannis-Meimarakis and G. Koutrika, “A survey on deep learning approaches for text-to-sql,” The VLDB Journal, 2023

    work page 2023

  52. [52]

    BERT: Pre-training of deep bidirectional transformers for language understanding,

    J. Devlin, M.-W. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of deep bidirectional transformers for language understanding,” in North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT), 2019

    work page 2019

  53. [53]

    TaBERT: Pretraining for joint understanding of textual and tabu- lar data

    P. Yin, G. Neubig, W.-t. Yih, and S. Riedel, “Tabert: Pretraining for joint understanding of textual and tabular data,” arXiv preprint arXiv:2005.08314 , 2020

    work page arXiv 2005

  54. [54]

    Ryan- sql: Recursively applying sketch-based slot fillings for complex text-to-sql in cross-domain databases,

    D. Choi, M. C. Shin, E. Kim, and D. R. Shin, “Ryan- sql: Recursively applying sketch-based slot fillings for complex text-to-sql in cross-domain databases,” Compu- tational Linguistics, 2021

    work page 2021