Pith. sign in

REVIEW 4 major objections 7 minor 46 references

Schema-R1: A reasoning training approach for schema linking in Text-to-SQL Task

T0 review · 4 major / 7 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Schema-R1 claims that a small language model trained for schema linking with 200 chain-of-thought samples plus GRPO reinforcement learning beats a fully supervised fine-tuning baseline on Spider-dev.

desk verdict A useful and plausible empirical extension of reasoning RL to schema linking, but Eq. (7) as written describes a reward that would reward missing tables, and that has to be fixed before the central claim is fully supported. read the letter →

arxiv 2506.11986 v1 pith:Y3ZVYTCY submitted 2025-06-13 cs.AI cs.CLcs.DB

classification cs.AIcs.CLcs.DB
keywords Text-to-SQLschemalinkingreasoningmodelreinforcementlearningGRPOchain-of-thoughtsupervisedfine-tuningSpiderbenchmark
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Schema-R1 claims that a schema-linking model for Text-to-SQL can be trained to reason instead of memorize by combining a tiny supervised cold start with reinforcement learning. The authors take 200 examples from Spider, prompt a large language model to write chain-of-thought explanations for the known correct answers, fine-tune a small open model on those examples, then train with GRPO using rule-based rewards for format, reasoning length, and schema-link accuracy. On Spider-dev the 1.5B model reaches 89.94 table FilteredAcc and 68.82 column FilteredAcc, beating the DTS-SQL fine-tuning baseline by about 15 and 26 percentage points respectively. This matters because schema linking is the step that tells a Text-to-SQL system which tables and columns to use, and the recipe needs only 200 labeled reasoning samples rather than a large supervised set.

What carries the argument

The load-bearing mechanism is a three-stage training pipeline: prompt-based chain-of-thought generation, cold-start supervised fine-tuning, and GRPO reinforcement learning. In the first stage, prompt templates containing the question, database schema, and ground-truth table/column sets are sent to DeepSeek-R1 to reconstruct the missing reasoning steps, yielding 200 chain-of-thought samples. Stage two is cold-start SFT on those samples, teaching the model to emit response sections for reasoning and for the final schema link. Stage three is GRPO, a group-relative policy optimization that draws multiple responses per query and computes advantages from three rule-based rewards: a format reward, a reasoning-length reward, and a schema-linking reward based on filter accuracy. The schema-linking reward is what ties the reinforcement signal directly to the task.

What would settle it

Look at the released code for the schema-linking reward: if it computes the reward exactly as written, using the set difference of ground truth minus prediction, the reported improvement should vanish; a reproduction run with that code would decide whether the equation or an overlap-based implementation produced the numbers.

Watch

Extended reading notes

Core claim

The paper's central claim is that decoupling schema linking from SQL generation and training a dedicated small language model with reasoning-oriented reinforcement learning outperforms conventional supervised fine-tuning. The evidence is the comparison on Spider-dev: cold-start SFT alone underperforms full SFT, but adding the GRPO stage lifts the 1.5B model to 73.21 exact match and 89.94 FilteredAcc for tables and 38.24 exact match and 68.82 FilteredAcc for columns, while the DTS-SQL baseline reaches 64.84 and 75.0 for tables and 31.17 and 42.43 for columns. The paper interprets the jump as a shift from rote learning to reflective reasoning: the model learns to generate thinking traces that lead to correct answer table and column sets.

Load-bearing premise

The central claim rests on the reward in Eq. (7) and Eq. (8) being implemented as an overlap between predicted and ground-truth table and column sets, not as the set difference the printed equations show, since a set difference would fail to reward correct predictions and the reported gains would not be possible.

Editorial extensions

If this is right

  • Cold-start SFT on 200 samples plus GRPO reasoning training exceeds full-data supervised fine-tuning on both table and column prediction, so small curated reasoning sets can substitute for large supervised sets.
  • Models trained this way produce explicit reasoning traces, making their schema-link decisions inspectable and eligible for rule-based rewards.
  • The 1.5B model benefits more from the reasoning stage than the 0.5B model, suggesting the approach scales with model size.
  • The same reward design covers both table prediction and column prediction, allowing a single model to output the full schema link.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The cold-start-plus-RL recipe separates the reasoning trace from the final answer, which should make schema-linking decisions easier to audit and correct in deployed text-to-SQL pipelines.
  • The same recipe may transfer to other structured-output prediction tasks where the answer is a set of schema elements, such as entity linking or feature selection.
  • Because the reward uses filter accuracy, the model optimizes coverage of correct schema elements; pairing the reward with an end-to-end SQL correctness signal could close the gap between schema linking and final query accuracy.
  • A direct next test would be running the same 200-sample cold start plus GRPO on other benchmarks such as BIRD, or on larger open models, to see whether the gains persist and grow with scale.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. The paper proposes Schema-R1, a three-stage training approach for schema linking in Text-to-SQL: (1) constructing 200 high-quality chain-of-thought samples via commercial LLMs (DeepSeek-R1), (2) supervised fine-tuning for cold-start initialization, and (3) GRPO-based reinforcement learning with rule-based rewards for table/column prediction. Experiments on Spider-dev with Qwen2.5-0.5B and Qwen2.5-1.5B report substantial improvements in table and column FilteredAcc over a DTS-SQL fine-tuned baseline, e.g., 89.94 vs. 75.0 for 1.5B table prediction. The paper claims at least a 10% improvement in filter accuracy and provides a GitHub repository.

Significance. If the results are reproducible, the work is a useful demonstration that reasoning-focused RL can improve a downstream NLP task like schema linking, even starting from a small cold-start SFT set. The approach is simple, uses open models, and targets a privacy-relevant application. The authors provide code, and the three-stage pipeline (prompt-based CoT generation, cold-start SFT, rule-based GRPO) is clearly a reasonable template for similar tasks. However, the manuscript as written contains a load-bearing notation error in the reward definition, omits key training hyperparameters, and reports single-run results without error bars, so the empirical claims are not yet fully substantiated.

major comments (4)
  1. [Sec. 4.2, Eq. (7)-(8)] The authors must clarify that the set difference in Eq. (7) is a typographical error and provide the correct reward formula. Without this correction, the method as described cannot work.
  2. [Sec. 5.1, training settings] The paper must include the missing reward and R L hyperparameters, as the reported gains are the core contribution.
  3. [Tables 1 & 2 and Sec. 5.1] The authors need to add error bars or multiple runs to support the quantitative claims.
  4. [Sec. 5.1, evaluation metrics] A precise, formal definition of FilteredAcc is needed; the current one-sentence description is insufficient.
minor comments (7)
  1. [Sec. 1] The acronym 'GPRO' appears in the introduction ('We employ GPRO with the SFTed model as reference'); it should be 'GRPO'.
  2. [Fig. 1 caption] The caption contains a typo: 'Rlue-Base' should be 'Rule-Based'.
  3. [Sec. 4.2] In the text following Eq. (7), 'the average reward for each correctly predicted item is calculated as Ptmax/len(t*_i)' should read 'Rtmax/len(t*_i)'. Using 'Ptmax' for the reward scale is confusing.
  4. [Sec. 5.1] The inference engine is written as 'VLLM'; the correct name is 'vLLM'.
  5. [Table 2 caption] The caption says 'Performance of table prediction for different method in Columns prediction Task'; it should say 'column prediction'.
  6. [Appendix A] In the prompt template, the XML-like tags are inconsistent: '<database>...<database>' should be closed as '</database>'.
  7. [Eq. (2)] The GRPO objective as typeset has mis-matched brackets and the KL penalty is written inside the expectation; a cleaner, complete formulation (including the definition of the advantage A-hat) would help reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the RL training optimizes a rule-based reward on the training split and evaluates on held-out Spider-dev; the Eq. (7) set-notation issue is a correctness flaw, not circularity.

full rationale

The paper's central claim is an empirical training result: Schema-R1 combines SFT on 200 CoT samples with GRPO rule-based RL and reports improved table/column FilteredAcc on Spider-dev. The reward signal is computed from ground-truth schema linking labels on the training split, and the evaluation metric is applied to the held-out dev split. This is standard reinforcement learning rather than circular reasoning: no parameter is fitted to the dev set and then reported as a prediction, and no derived quantity is definitionally equal to its input. The method's comparison is against DTS-SQL, an external fine-tuning baseline, and the citations used for task construction and evaluation metrics are not authored by the present authors, so there is no load-bearing self-citation chain. The most notable concern in the paper, Eq. (7), writes the 'correctly predicted table set' as t*_i - t'_i, which in set notation is the set of missed ground-truth tables rather than the overlap t*_i ∩ t'_i; if implemented literally, this would penalize omissions instead of rewarding hits and could not explain the reported gains. However, this is an internal consistency and reproducibility issue, not a circularity: it does not make the reported evaluation equivalent to the training objective by construction, and it does not involve a fitted input being renamed as a prediction. The paper is therefore self-contained against external benchmarks, and no circular step can be exhibited from the text.

Assumptions & free parameters 8 free parameters · 5 assumptions · 0 invented entities

The paper's central claim is empirical and rests on a standard GRPO pipeline. The main unpaid inputs are hand-chosen reward weights and length bounds, which are not reported, plus domain assumptions about whether DeepSeek-R1-generated CoTs and filter-accuracy overlap are appropriate training signals.

free parameters (8)
  • Rtmax (maximum table reward) = not reported
    Hand-chosen reward weight in Eq. (7); the paper states only Rtmax > Rcmax, so the exact value affects RL behavior but is not given.
  • Ptmax (maximum table penalty) = not reported
    Hand-chosen in Eq. (7); not specified numerically.
  • Rcmax (maximum column reward) = not reported
    Hand-chosen in Eq. (8); values not reported.
  • Pcmax (maximum column penalty) = not reported
    Hand-chosen in Eq. (8); values not reported.
  • Lower Length and Upper Length (reasoning length bounds) = not reported
    Hand-chosen boundaries in Eq. (6) for the reasoning length reward; values are not given.
  • KL penalty coefficient beta (GRPO) = not reported
    Standard GRPO hyperparameter; not specified in Sec. 5.1.
  • Number of RL steps or epochs = not reported
    Training curves show 70 steps but no explicit total number of GRPO steps or convergence criterion.
  • Cold-start CoT sample count = 200
    Hand-chosen subset size for the SFT cold start; no selection criteria are given.
assumptions (5)
  • domain assumption Ground-truth tables and columns extracted from SQL queries provide a valid supervision signal for schema linking.
    Adopted from [11] and used to construct training data and rewards; if extraction is noisy, all reported metrics inherit that noise.
  • domain assumption DeepSeek-R1-generated chain-of-thought explanations, prompted with ground-truth answers, are high-quality reasoning samples.
    The cold-start SFT relies on these CoTs; the paper does not evaluate their quality beyond final task accuracy.
  • standard math The GRPO objective from DeepSeekMath is a valid and stable reasoning RL update.
    The method uses the published GRPO objective in Eq. (2) without modification.
  • domain assumption Schema linking can be evaluated by filter accuracy, the overlap between predicted and gold table and column sets.
    FilteredAcc is used both as evaluation metric and as reward signal; this assumes set overlap is the right objective.
  • ad hoc to paper Table reward should exceed column reward because table errors cascade to columns.
    This hierarchy is asserted in Sec. 4.2 without empirical justification; if wrong, the RL update is misaligned.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Schema-R1: A reasoning training approach for schema linking in Text-to-SQL Task." pith.science (2026). https://pith.science/paper/Y3ZVYTCY

@misc{pith2026250611986,
  author       = {Pith},
  title        = {Pith review of: Schema-R1: A reasoning training approach for schema linking in Text-to-SQL Task},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Y3ZVYTCY}},
  note         = {Machine review of arXiv:2506.11986}
}
read the original abstract

Schema linking is a critical step in Text-to-SQL task, aiming to accurately predict the table names and column names required for the SQL query based on the given question. However, current fine-tuning approaches for schema linking models employ a rote-learning paradigm, excessively optimizing for ground truth schema linking outcomes while compromising reasoning ability. This limitation arises because of the difficulty in acquiring a high-quality reasoning sample for downstream tasks. To address this, we propose Schema-R1, a reasoning schema linking model trained using reinforcement learning. Specifically, Schema-R1 consists of three key steps: constructing small batches of high-quality reasoning samples, supervised fine-tuning for cold-start initialization, and rule-based reinforcement learning training. The final results demonstrate that our method effectively enhances the reasoning ability of the schema linking model, achieving a 10\% improvement in filter accuracy compared to the existing method. Our code is available at https://github.com/hongWin/Schema-R1/.

Figures

Figures reproduced from arXiv: 2506.11986 by the authors.

Figure 1
Figure 1. The difference between SFT and Reasoning RL in Schema linking downstream task tuning [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. The training process of schema-R1 on different base models [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Prompt template for CoT generation 11 [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

46 extracted references · 22 canonical work pages

  1. [1]

    A survey on employing large language models for text-to-sql tasks

    Liang Shi, Zhengju Tang, Nan Zhang, Xiaotong Zhang, and Zhi Yang. A survey on employing large language models for text-to-sql tasks. ACM Computing Surveys, 2024

  2. [2]

    Sql-r1: Training natural language to sql reasoning model by reinforcement learning

    Peixian Ma, Xialie Zhuang, Chengjin Xu, Xuhui Jiang, Ran Chen, and Jian Guo. Sql-r1: Training natural language to sql reasoning model by reinforcement learning. arXiv preprint arXiv:2504.08600, 2025

  3. [3]

    Re-examining the role of schema linking in text-to-SQL

    Wenqiang Lei, Weixin Wang, Zhixin Ma, Tian Gan, Wei Lu, Min-Yen Kan, and Tat-Seng Chua. Re-examining the role of schema linking in text-to-SQL. In Bonnie Webber, Trevor Cohn, Yulan He, and Yang Liu, editors, Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 6943–6954, Online, November 2020. Association f...

  4. [4]

    Text-to-sql empowered by large language models: A benchmark evaluation

    Dawei Gao, Haibin Wang, Yaliang Li, Xiuyu Sun, Yichen Qian, Bolin Ding, and Jingren Zhou. Text-to-sql empowered by large language models: A benchmark evaluation. Proc. VLDB Endow., 17(5):1132–1145, January 2024

  5. [5]

    Demonstration of db-gpt: Next generation data interaction system empowered by large language models

    Siqiao Xue, Danrui Qi, Caigao Jiang, Fangyin Cheng, Keting Chen, Zhiping Zhang, Hongyang Zhang, Ganglin Wei, Wang Zhao, Fan Zhou, Hong Yi, Shaodong Liu, Hongjun Yang, and Faqiang Chen. Demonstration of db-gpt: Next generation data interaction system empowered by large language models. 17(12), 2024

  6. [6]

    Open-sql framework: Enhancing text-to-sql on open-source large language models

    Xiaojun Chen, Tianle Wang, Tianhao Qiu, Jianbin Qin, and Min Yang. Open-sql framework: Enhancing text-to-sql on open-source large language models. arXiv preprint arXiv:2405.06674, 2024

  7. [7]

    Sean Wang

    Yuankai Fan, Zhenying He, Tonghui Ren, Can Huang, Yinan Jing, Kai Zhang, and X. Sean Wang. Metasql: A generate-then-rank framework for natural language to sql translation. In 2024 IEEE 40th International Conference on Data Engineering (ICDE), pages 1765–1778, 2024

  8. [8]

    Din-sql: decomposed in-context learning of text-to-sql with self-correction

    Mohammadreza Pourreza and Davood Rafiei. Din-sql: decomposed in-context learning of text-to-sql with self-correction. In Proceedings of the 37th International Conference on Neural Information Processing Systems, NIPS ’23, Red Hook, NY , USA, 2023. Curran Associates Inc

Show all 46 references
  1. [9]

    MCS-SQL: Leveraging multiple prompts and multiple-choice selection for text-to-SQL generation

    Dongjun Lee, Choongwon Park, Jaehyuk Kim, and Heesoo Park. MCS-SQL: Leveraging multiple prompts and multiple-choice selection for text-to-SQL generation. In Owen Rambow, Leo Wanner, Marianna Apidianaki, Hend Al-Khalifa, Barbara Di Eugenio, and Steven Schockaert, editors, Proce...

  2. [10]

    Act-sql: In-context learning for text-to-sql with automatically-generated chain-of-thought

    Hanchong Zhang, Ruisheng Cao, Lu Chen, Hongshen Xu, and Kai Yu. Act-sql: In-context learning for text-to-sql with automatically-generated chain-of-thought. In EMNLP (Findings), 2023

  3. [11]

    DTS-SQL: Decomposed text-to-SQL with small large language models

    Mohammadreza Pourreza and Davood Rafiei. DTS-SQL: Decomposed text-to-SQL with small large language models. In Yaser Al-Onaizan, Mohit Bansal, and Yun-Nung Chen, editors, Findings of the Association for Computational Linguistics: EMNLP 2024, pages 8212–8220, Miami, Florida, USA...

  4. [12]

    Instruction tuning text-to-sql with large language models in the power grid domain

    Gang Sun, Ran Shen, Liangfeng Jin, Yifan Wang, Shiyu Xu, Jinpeng Chen, and Weihao Jiang. Instruction tuning text-to-sql with large language models in the power grid domain. In Proceedings of the 2023 4th International Conference on Control, Robotics and Intelligent System, pag...

  5. [13]

    Msc-sql: Multi-sample critiquing small language models for text-to-sql translation

    Satya Krishna Gorti, Ilan Gofman, Zhaoyan Liu, Jiapeng Wu, NoÃG, l V ouitsis, Guangwei Yu, Jesse C Cresswell, and Rasa Hosseinzadeh. Msc-sql: Multi-sample critiquing small language models for text-to-sql translation. arXiv preprint arXiv:2410.12916, 2024

  6. [14]

    Datagpt-sql-7b: An open-source language model for text-to-sql

    Lixia Wu, Peng Li, Junhong Lou, and Lei Fu. Datagpt-sql-7b: An open-source language model for text-to-sql. arXiv preprint arXiv:2409.15985, 2024

  7. [15]

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

    Haoyang Li, Jing Zhang, Hanbing Liu, Ju Fan, Xiaokang Zhang, Jun Zhu, Renjie Wei, Hongyan Pan, Cuiping Li, and Hong Chen. Codes: Towards building open-source language models for text-to-sql. Proceedings of the ACM on Management of Data, 2(3):1–28, 2024

  8. [16]

    Cogsql: A cognitive framework for enhancing large language models in text-to-sql translation

    Hongwei Yuan, Xiu Tang, Ke Chen, Lidan Shou, Gang Chen, and Huan Li. Cogsql: A cognitive framework for enhancing large language models in text-to-sql translation. Proceedings of the AAAI Conference on Artificial Intelligence, 39(24):25778–25786, Apr. 2025

  9. [17]

    Chain-of-thought prompting elicits reasoning in large language models.Advances in neural information processing systems, 35:24824–24837, 2022

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. Chain-of-thought prompting elicits reasoning in large language models.Advances in neural information processing systems, 35:24824–24837, 2022. 8 Preprint PRIME AI paper

  10. [18]

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948, 2025

  11. [19]

    Deepseekmath: Pushing the limits of mathematical reasoning in open language models

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Y Wu, et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300, 2024

  12. [20]

    Reinforce++: A simple and efficient approach for aligning large language models

    Jian Hu. Reinforce++: A simple and efficient approach for aligning large language models. arXiv preprint arXiv:2501.03262, 2025

  13. [21]

    Dapo: An open-source llm reinforcement learning system at scale

    Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Tiantian Fan, Gaohong Liu, Lingjun Liu, Xin Liu, et al. Dapo: An open-source llm reinforcement learning system at scale. arXiv preprint arXiv:2503.14476, 2025

  14. [22]

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

    Haoyang Li, Jing Zhang, Cuiping Li, and Hong Chen. Resdsql: decoupling schema linking and skeleton parsing for text-to-sql. In Proceedings of the Thirty-Seventh AAAI Conference on Artificial Intelligence and Thirty-Fifth Conference on Innovative Applications of Artificial Inte...

  15. [23]

    Catsql: Towards real world natural language to sql applications

    Han Fu, Chang Liu, Bin Wu, Feifei Li, Jian Tan, and Jianling Sun. Catsql: Towards real world natural language to sql applications. Proceedings of the VLDB Endowment, 16(6):1534–1547, 2023

  16. [24]

    RAT-SQL: Relation- aware schema encoding and linking for text-to-SQL parsers

    Bailin Wang, Richard Shin, Xiaodong Liu, Oleksandr Polozov, and Matthew Richardson. RAT-SQL: Relation- aware schema encoding and linking for text-to-SQL parsers. In Dan Jurafsky, Joyce Chai, Natalie Schluter, and Joel Tetreault, editors, Proceedings of the 58th Annual Meeting ...

  17. [25]

    LGESQL: Line graph enhanced text-to-SQL model with mixed local and non-local relations

    Ruisheng Cao, Lu Chen, Zhi Chen, Yanbin Zhao, Su Zhu, and Kai Yu. LGESQL: Line graph enhanced text-to-SQL model with mixed local and non-local relations. In Chengqing Zong, Fei Xia, Wenjie Li, and Roberto Navigli, editors, Proceedings of the 59th Annual Meeting of the Associat...

  18. [26]

    C3: Zero-shot text-to-sql with chatgpt

    Xuemei Dong, Chao Zhang, Yuhang Ge, Yuren Mao, Yunjun Gao, Jinshu Lin, Dongfang Lou, et al. C3: Zero-shot text-to-sql with chatgpt. arXiv preprint arXiv:2307.07306, 2023

  19. [27]

    Enhancing text-to-SQL capabilities of large language models through tailored promptings

    Zhao Tan, Xiping Liu, Qing Shu, Xi Li, Changxuan Wan, Dexi Liu, Qizhi Wan, and Guoqiong Liao. Enhancing text-to-SQL capabilities of large language models through tailored promptings. In Nicoletta Calzolari, Min-Yen Kan, Veronique Hoste, Alessandro Lenci, Sakriani Sakti, and Ni...

  20. [28]

    Pet-sql: A prompt-enhanced two-round refinement of text-to-sql with cross-consistency

    Zhishuai Li, Xiang Wang, Jingjing Zhao, Sun Yang, Guoqing Du, Xiaoru Hu, Bin Zhang, Yuxiao Ye, Ziyue Li, Rui Zhao, et al. Pet-sql: A prompt-enhanced two-round refinement of text-to-sql with cross-consistency. arXiv preprint arXiv:2403.09732, 2024

  21. [29]

    Sql-to-schema enhances schema linking in text-to-sql

    Sun Yang, Qiong Su, Zhishuai Li, Ziyue Li, Hangyu Mao, Chenxi Liu, and Rui Zhao. Sql-to-schema enhances schema linking in text-to-sql. In Christine Strauss, Toshiyuki Amagasa, Giuseppe Manco, Gabriele Kotsis, A. Min Tjoa, and Ismail Khalil, editors, Database and Expert Systems...

  22. [30]

    Spsql: Step-by-step parsing based framework for text-to-sql generation

    Ran Shen, Gang Sun, Hao Shen, Yiling Li, Liangfeng Jin, and Han Jiang. Spsql: Step-by-step parsing based framework for text-to-sql generation. In 2023 7th International Conference on Machine Vision and Information Technology (CMVIT), pages 115–122. IEEE, 2023

  23. [31]

    Tablellm: Enabling tabular data manipulation by llms in real office usage scenarios

    Xiaokang Zhang, Sijia Luo, Bohan Zhang, Zeyao Ma, Jing Zhang, Yang Li, Guanlin Li, Zijun Yao, Kangli Xu, Jinchang Zhou, et al. Tablellm: Enabling tabular data manipulation by llms in real office usage scenarios. arXiv preprint arXiv:2403.19318, 2024

  24. [32]

    Tree of thoughts: Deliberate problem solving with large language models

    Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. Tree of thoughts: Deliberate problem solving with large language models. Advances in neural information processing systems, 36:11809–11822, 2023

  25. [33]

    Graph chain-of-thought: Augmenting large language models by reasoning on graphs

    Bowen Jin, Chulin Xie, Jiawei Zhang, Kashob Kumar Roy, Yu Zhang, Zheng Li, Ruirui Li, Xianfeng Tang, Suhang Wang, Yu Meng, and Jiawei Han. Graph chain-of-thought: Augmenting large language models by reasoning on graphs. In Lun-Wei Ku, Andre Martins, and Vivek Srikumar, editors...

  26. [34]

    Thinkless: Llm learns when to think

    Gongfan Fang, Xinyin Ma, and Xinchao Wang. Thinkless: Llm learns when to think. arXiv preprint arXiv:2505.13379, 2025

  27. [35]

    Think only when you need with large hybrid-reasoning models

    Lingjie Jiang, Xun Wu, Shaohan Huang, Qingxiu Dong, Zewen Chi, Li Dong, Xingxing Zhang, Tengchao Lv, Lei Cui, and Furu Wei. Think only when you need with large hybrid-reasoning models. arXiv preprint arXiv:2505.14631, 2025

  28. [36]

    Reasoning models can be effective without thinking

    Wenjie Ma, Jingxuan He, Charlie Snell, Tyler Griggs, Sewon Min, and Matei Zaharia. Reasoning models can be effective without thinking. arXiv preprint arXiv:2504.09858, 2025

  29. [37]

    Acemath: Advancing frontier math reasoning with post-training and reward modeling

    Zihan Liu, Yang Chen, Mohammad Shoeybi, Bryan Catanzaro, and Wei Ping. Acemath: Advancing frontier math reasoning with post-training and reward modeling. arXiv preprint, 2024

  30. [38]

    Light-r1: Curriculum sft, dpo and rl for long cot from scratch and beyond

    Liang Wen, Yunke Cai, Fenrui Xiao, Xin He, Qi An, Zhenyu Duan, Yimin Du, Junchen Liu, Lifu Tang, Xiaowei Lv, et al. Light-r1: Curriculum sft, dpo and rl for long cot from scratch and beyond. arXiv preprint arXiv:2503.10460, 2025

  31. [39]

    Rm-r1: Reward modeling as reasoning

    Xiusi Chen, Gaotang Li, Ziqi Wang, Bowen Jin, Cheng Qian, Yu Wang, Hongru Wang, Yu Zhang, Denghui Zhang, Tong Zhang, et al. Rm-r1: Reward modeling as reasoning. arXiv preprint arXiv:2505.02387, 2025

  32. [40]

    Inference-time scaling for generalist reward modeling

    Zijun Liu, Peiyi Wang, Runxin Xu, Shirong Ma, Chong Ruan, Peng Li, Yang Liu, and Yu Wu. Inference-time scaling for generalist reward modeling. arXiv preprint arXiv:2504.02495, 2025

  33. [41]

    Fin-r1: A large language model for financial reasoning through reinforcement learning

    Zhaowei Liu, Xin Guo, Fangqi Lou, Lingfeng Zeng, Jinyi Niu, Zixuan Wang, Jiajie Xu, Weige Cai, Ziwei Yang, Xueqian Zhao, et al. Fin-r1: A large language model for financial reasoning through reinforcement learning. arXiv preprint arXiv:2503.16252, 2025

  34. [42]

    Table-r1: Inference-time scaling for table reasoning

    Zheyuan Yang, Lyuhao Chen, Arman Cohan, and Yilun Zhao. Table-r1: Inference-time scaling for table reasoning. arXiv preprint arXiv:2505.23621, 2025

  35. [43]

    Reasoning-table: Exploring reinforcement learning for table reasoning

    Fangyu Lei, Jinxiang Meng, Yiming Huang, Tinghong Chen, Yun Zhang, Shizhu He, Jun Zhao, and Kang Liu. Reasoning-table: Exploring reinforcement learning for table reasoning. arXiv preprint arXiv:2506.01710, 2025

  36. [44]

    Proximal policy optimization algorithms

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347, 2017

  37. [45]

    L1: Controlling how long a reasoning model thinks with reinforcement learning

    Pranjal Aggarwal and Sean Welleck. L1: Controlling how long a reasoning model thinks with reinforcement learning. arXiv preprint arXiv:2503.04697, 2025

  38. [46]

    Thinkprune: Pruning long chain-of-thought of llms via reinforcement learning

    Bairu Hou, Yang Zhang, Jiabao Ji, Yujian Liu, Kaizhi Qian, Jacob Andreas, and Shiyu Chang. Thinkprune: Pruning long chain-of-thought of llms via reinforcement learning. arXiv preprint arXiv:2504.01296, 2025. 10 Preprint PRIME AI paper A Prompt Construction CoT generation promp...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.