-
Notifications
You must be signed in to change notification settings - Fork 16
Expand file tree
/
Copy pathindex.html
More file actions
519 lines (471 loc) · 30.2 KB
/
index.html
File metadata and controls
519 lines (471 loc) · 30.2 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<!-- Meta tags for social media banners, these should be filled in appropriatly as they are your "business card" -->
<!-- Replace the content tag with appropriate information -->
<meta name="description" content="SQL-R1: Training Natural Language to SQL Reasoning Model By Reinforcement Learning">
<meta property="og:title" content="SQL-R1"/>
<meta property="og:description" content="Training Natural Language to SQL Reasoning Model By Reinforcement Learning"/>
<meta property="og:url" content="https://github.com/MPX0222/SQL-R1"/>
<!-- Path to banner image, should be in the path listed below. Optimal dimenssions are 1200X630-->
<meta property="og:image" content="images/overview.png" />
<meta property="og:image:width" content="1200"/>
<meta property="og:image:height" content="630"/>
<meta name="twitter:title" content="TWITTER BANNER TITLE META TAG">
<meta name="twitter:description" content="TWITTER BANNER DESCRIPTION META TAG">
<!-- Path to banner image, should be in the path listed below. Optimal dimenssions are 1200X600-->
<meta name="twitter:image" content="static/images/your_twitter_banner_image.png">
<meta name="twitter:card" content="summary_large_image">
<!-- Keywords for your paper to be indexed by-->
<meta name="keywords" content="KEYWORDS SHOULD BE PLACED HERE">
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta name="citation_title" content="SQL-R1: Training Natural Language to SQL Reasoning Model by Reinforcement Learning">
<meta name="citation_arxiv_id" content="2504.08600">
<meta name="citation_author" content="Peixian Ma">
<meta name="citation_author" content="Xialie Zhuang">
<meta name="citation_author" content="Chengjin Xu">
<meta name="citation_author" content="Xuhui Jiang">
<meta name="citation_author" content="Ran Chen">
<meta name="citation_author" content="Jian Guo">
<meta name="citation_publication_date" content="2025/04/11">
<title>SQL-R1: Training Natural Language to SQL Reasoning Model By Reinforcement Learning</title>
<!-- <link rel="icon" type="image/x-icon" href="static/images/favicon.ico"> -->
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro"
rel="stylesheet">
<link rel="stylesheet" href="static/css/bulma.min.css">
<link rel="stylesheet" href="static/css/bulma-carousel.min.css">
<link rel="stylesheet" href="static/css/bulma-slider.min.css">
<link rel="stylesheet" href="static/css/fontawesome.all.min.css">
<link rel="stylesheet"
href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="static/css/index.css">
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script src="https://documentcloud.adobe.com/view-sdk/main.js"></script>
<script defer src="static/js/fontawesome.all.min.js"></script>
<script src="static/js/bulma-carousel.min.js"></script>
<script src="static/js/bulma-slider.min.js"></script>
<script src="static/js/index.js"></script>
<!-- MathJax for mathematical formulas -->
<script src="https://polyfill.io/v3/polyfill.min.js?features=es6"></script>
<script id="MathJax-script" async src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js"></script>
<script>
window.MathJax = {
tex: {
inlineMath: [['$', '$'], ['\\(', '\\)']],
displayMath: [['$$', '$$'], ['\\[', '\\]']]
}
};
</script>
</head>
<body>
<!-- Header with Conference and Institution Logos -->
<section class="header-logos">
<div class="container is-max-desktop">
<div class="columns is-centered is-vcentered">
<div class="column is-narrow">
<div class="logo-container">
<!-- Conference Logo (NeurIPS 2025) -->
<div class="conference-logo">
<img src="static/images/NeurIPS-logo.svg" alt="NeurIPS 2025" style="height: 80px;"/>
</div>
</div>
</div>
<div class="column is-narrow">
<div class="logo-separator" style="width: 1px; height: 40px; background-color: #dbdbdb; margin: 0 2rem;"></div>
</div>
<div class="column is-narrow">
<div class="logo-container">
<!-- Institution Logos -->
<div class="institution-logos" style="display: flex; align-items: center; padding-top: 10px;">
<!-- IDEA Research Logo -->
<div class="institution-logo">
<img src="static/images/idea_logo.png" alt="IDEA Research" style="height: 80px;"/>
</div>
<!-- HKUST(GZ) Logo -->
<div class="institution-logo">
<img src="static/images/hkust_logo.png" alt="HKUST(GZ)" style="height: 50px; padding-left: 30px;"/>
</div>
<div class="institution-logo">
<img src="static/images/ucas_logo.png" alt="UCAS" style="height: 50px; padding-left: 30px;"/>
</div>
<!-- Other institutions can be added here -->
<div class="institution-logo">
<img src="static/images/dataarc_logo.png" alt="DataArc" style="height: 100px; padding-left: 30px;"/>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<h1 class="title is-1 publication-title">SQL-R1: Training Natural Language to SQL Reasoning Model By Reinforcement Learning</h1>
<div class="is-size-5 publication-authors">
<span class="author-block">
<a href="https://github.com/MPX0222" target="_blank">Peixian Ma</a><sup>1,2</sup>,</span>
<span class="author-block">
<a href="#" target="_blank">Xialie Zhuang</a><sup>1,3</sup>,</span>
<span class="author-block">
<a href="#" target="_blank">Chengjin Xu</a><sup>1,4</sup>,</span>
<span class="author-block">
<a href="#" target="_blank">Xuhui Jiang</a><sup>1,4</sup>,</span>
<span class="author-block">
<a href="#" target="_blank">Ran Chen</a><sup>1</sup>,</span>
<span class="author-block">
<a href="#" target="_blank">Jian Guo</a><sup>1</sup>
</span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block">
<sup>1</sup>IDEA Research, International Digital Economy Academy, <br>
<sup>2</sup>The Hong Kong University of Science and Technology (Guangzhou), <br>
<sup>3</sup>University of Chinese Academy of Science, <br>
<sup>4</sup>DataArc Tech Ltd.
</span>
<!-- <span class="eql-cntrb"><small><br><sup>*</sup>Indicates Equal Contribution</small></span> -->
</div>
<span class="author-block" style="color: rgb(128, 0, 128); font-weight: bold; font-size: 1.5em;">🎉 Accepted by NeurIPS 2025 Poster 🎉</span>
<div class="column has-text-centered">
<div class="publication-links">
<span class="link-block">
<a href="https://arxiv.org/pdf/2504.08600.pdf" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>Paper</span>
</a>
</span>
<span class="link-block">
<a href="https://github.com/MPX0222/SQL-R1" target="_blank"
class="external-ink button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code</span>
</a>
</span>
<span class="link-block">
<a href="https://huggingface.co/MPX0222forHF/SQL-R1-7B" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
🤗
</span>
<span>HuggingFace</span>
</a>
</span>
<span class="link-block">
<a href="https://modelscope.cn/models/MPX0222/SQL-R1-7B" target="_blank"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
🤖
</span>
<span>ModelScope</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- Introduction Section -->
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Introduction</h2>
<div class="content has-text-justified" style="line-height: 1.7; color: #2c3e50;">
<p style="margin-bottom: 1.5rem;">
Natural Language to SQL (NL2SQL) enables intuitive interactions with databases by transforming natural language queries into structured SQL statements. Recent advancements in NL2SQL have significantly enhanced the level of human-computer interaction within database query applications and contribute to a wide range of data science analysis tasks. Current NL2SQL models mainly focus on optimizing workflows and their components, such as <em>schema linking</em>, <em>content retrieval</em>, and <em>generation correction</em>.
</p>
<p style="margin-bottom: 1.5rem;">
Despite these advancements, <strong style="color: #3273dc;">significant challenges persist</strong>, particularly regarding the inference performance in complex scenarios involving multi-table joins and nested queries. Current methodologies primarily utilize supervised fine-tuning (SFT) to train NL2SQL models, which may limit adaptability and interpretability in new environments (e.g., finance and healthcare).
</p>
<p style="margin-bottom: 1.5rem;">
In order to enhance the reasoning performance of NL2SQL models in complex situations, we introduce <strong style="color: #3273dc;">SQL-R1</strong>, a novel NL2SQL reasoning model trained by reinforcement learning (RL) algorithms. We design a specialized RL-based reward function tailored for NL2SQL tasks and discuss the impact of cold start on the effectiveness of intensive training. Additionally, we achieve competitive accuracy using only a tiny amount of synthetic NL2SQL data for augmented training and further explore data engineering for RL.
</p>
<div style="background: linear-gradient(135deg, #667eea15 0%, #764ba215 100%); border-left: 4px solid #3273dc; padding: 1.5rem; margin: 2rem 0; border-radius: 6px;">
<p style="margin-bottom: 0; color: #2c3e50;">
<strong>Key Results:</strong> SQL-R1 achieves execution accuracy of <strong style="color: #3273dc;">88.6%</strong> on the Spider benchmark and <strong style="color: #3273dc;">67.1%</strong> on the BIRD benchmark, demonstrating state-of-the-art performance in natural language to SQL translation tasks.
</p>
</div>
<div class="columns is-centered has-text-centered" style="margin-top: 2.5rem;">
<div class="column">
<img src="images/overview.png" alt="SQL-R1 Overview" style="width: 100%; height: auto; border-radius: 8px; box-shadow: 0 4px 12px rgba(0,0,0,0.1);"/>
<p class="has-text-grey" style="margin-top: 1.2rem; font-size: 0.95rem; font-style: italic;">
<strong>Figure 1:</strong> Overview of SQL-R1 training framework using reinforcement learning for NL2SQL tasks.
</p>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- Method Section -->
<section class="section" style="background-color: #fafafa;">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Method</h2>
<div class="content has-text-justified">
<div class="method-content">
<p>
This section introduces two forms of training NL2SQL models via RL algorithms: direct reinforcement training and reinforcement training via cold start after SFT. Cold start refers to using specific data to train the base model by SFT first so that it has a particular ability to think and follow instructions. In addition, due to the limited real data, we use the latest synthetic data to support the training process.
</p>
<!-- Data Preparation -->
<div class="method-subsection" style="margin-top: 2rem;">
<h3 class="title is-4">Data Preparation</h3>
<p>
We utilize the SynSQL-2.5M dataset as our primary data source, which is the first million-scale synthetic NL2SQL dataset, encompassing over 2.5 million diverse and high-quality data samples. The dataset features more than 16,000 synthetic databases across various domains, ensuring extensive coverage of realistic scenarios.
</p>
<p>
<strong>SFT Dataset:</strong> We investigated the impact of cold start condition in SFT on RL training. We utilized a dataset comprising 200,000 samples drawn from SynSQL-2.5M for SFT training (<strong>SynSQL-200K</strong>), with uniform sample size across different difficulty levels. For each sample $v = (x, t, y^*)$, $x$ represents the NL, $t$ represents the reasoning process enclosed in <code><think>...</think></code> tags, and $y^*$ denotes the SQL enclosed in <code><answer>...</answer></code> tags.
</p>
<p>
<strong>RL Dataset:</strong> We randomly sampled 5K NL-SQL pairs from SynSQL-2.5M with <strong>Complex</strong> difficulty (<strong>SynSQL-Complex-5K</strong>). For each pair $v = (x, y^*)$ in the RL dataset, $x$ represents the NL while $y^*$ denotes the SQL candidate. The aim of reinforcement learning is to enhance the accuracy of the answers and ensure they adhere to the expected format.
</p>
</div>
<!-- Reinforcement Training -->
<div class="method-subsection" style="margin-top: 2rem;">
<h3 class="title is-4">Reinforcement Training with GRPO</h3>
<p>
In the reinforcement learning phase, we employ the Group Relative Policy Optimization (GRPO) algorithm to enhance our training protocol, which obviates the need for the value model and operates with less memory requirements. For each natural language question aligned with its corresponding database schema, the policy model generates a set of $G$ SQL candidates $\{o_1, o_2, \ldots, o_G\}$ from the old policy $\pi_{old}$, which are evaluated using a composite reward function.
</p>
<div class="has-text-centered" style="margin: 1.5rem 0;">
<p>$$\mathcal{J}_{\text{GRPO}}(\theta) = \mathbb{E}_{\mathbf{v} \sim P(\mathbf{V}), \{o_i\}_{i=1}^G \sim \pi_{\theta_{\text{old}}}(O|\mathbf{v})} \left[ \frac{1}{G} \sum_{i=1}^G \left( \min \left( r_i^{\text{ratio}} A_i, \text{clip} \left( r_i^{\text{ratio}}, 1-\epsilon, 1+\epsilon \right) A_i \right) - \beta D_{\text{KL}}(\pi_\theta \| \pi_{\text{ref}}) \right) \right]$$</p>
</div>
<p>
where $r_i^{\text{ratio}} = \frac{\pi_\theta(o_i | V)}{\pi_{old}(o_i | V)}$ represents the importance sampling ratio; $A_i$ represents the group-relative advantage for each output; the clipping operator, hyperparameter $\epsilon$ and $\beta$ control the update step and divergence regularization; $\pi_{\text{ref}}$ represents the reference policy.
</p>
</div>
<!-- Reward Function Design -->
<div class="method-subsection" style="margin-top: 2rem;">
<h3 class="title is-4">Reward Function Design</h3>
<p>
We utilize a progressive feedback mechanism consisting of four types of rewards: Format Reward, Execution Reward, Result Reward, and Length Reward. This layered approach enhances the model's learning by providing detailed feedback at various stages.
</p>
<p style="margin-top: 1rem;">
<strong>Format Reward ($S_f$):</strong> We encourage the model to enclose the reasoning process within <code><think>...</think></code> tags and the final answer within <code><answer>...</answer></code> tags. SQL statements must be contained within <code>```sql...```</code> tags.
</p>
<div class="has-text-centered" style="margin: 1rem 0;">
<p>$$S_f = \begin{cases} 1, & \text{if format is correct} \\ -1, & \text{if format is incorrect} \end{cases}$$</p>
</div>
<p style="margin-top: 1rem;">
<strong>Execution Reward ($S_e$):</strong> Evaluates the syntactic correctness of SQL candidates, preventing the model from generating unexecutable responses.
</p>
<div class="has-text-centered" style="margin: 1rem 0;">
<p>$$S_e = \begin{cases} 2, & \text{if SQL candidate is executable} \\ 0, & \text{if format is incorrect} \\ -2, & \text{if SQL candidate is not executable} \end{cases}$$</p>
</div>
<p style="margin-top: 1rem;">
<strong>Result Reward ($S_r$):</strong> Evaluates the accuracy of query results using Execution Accuracy (EX).
</p>
<div class="has-text-centered" style="margin: 1rem 0;">
<p>$$S_r = \begin{cases} 3, & \text{if query result is correct} \\ 0, & \text{if format is incorrect or not executable} \\ -3, & \text{if query result is incorrect} \end{cases}$$</p>
</div>
<p style="margin-top: 1rem;">
<strong>Length Reward ($S_l$):</strong> Incentivizes the model to produce comprehensive reasoning processes while avoiding superfluous explanations.
</p>
<div class="has-text-centered" style="margin: 1rem 0;">
<p>$$S_l = \begin{cases} 0.5 \times S_{tl} + S_{al}, & \text{if correct and } len_{response} \leq \text{MAX\_LENGTH} \\ 0.5 + S_{al}, & \text{if correct and } len_{response} > \text{MAX\_LENGTH} \\ 0, & \text{other cases} \end{cases}$$</p>
</div>
<p style="font-size: 0.95rem; color: #666;">
where $S_{tl} = (len_{think} + len_{answer}) / \text{MAX\_LENGTH}$ and $S_{al} = len_{sql} / len_{answer}$.
</p>
</div>
<!-- SQL Candidates Selection -->
<div class="method-subsection" style="margin-top: 2rem;">
<h3 class="title is-4">SQL Candidates Selection</h3>
<p>
To select the most appropriate SQL in the reasoning process, the model generates several SQL candidates and their thought processes for a problem. We execute all SQL candidates and select the SQL with the highest score as the final answer based on self-consistency voting. Notably, the reasoning response of SQL-R1 comprises an observable process of thinking and interpreting, making the results easier for the user to understand.
</p>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- Experiment Section -->
<section class="section">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Experiments</h2>
<div class="content has-text-justified">
<div class="experiment-content">
<p>
We conduct comprehensive experiments on two widely-used NL2SQL benchmarks: <strong>Spider</strong> and <strong>BIRD</strong>. These datasets cover a wide range of database schemas and query complexities, providing a robust evaluation of our model's capabilities.
</p>
<!-- Main Results Subsection -->
<div class="experiment-subsection" style="margin-top: 2rem;">
<h3 class="title is-4">Main Results</h3>
<p>
Our SQL-R1 model demonstrates state-of-the-art performance on both benchmark datasets. The results show significant improvements over baseline supervised fine-tuning approaches:
</p>
<ul style="margin: 1rem 0;">
<li><strong>Spider Benchmark:</strong> SQL-R1 achieves an execution accuracy of <strong>88.6%</strong>, outperforming previous state-of-the-art methods</li>
<li><strong>BIRD Benchmark:</strong> SQL-R1 achieves an execution accuracy of <strong>67.1%</strong>, demonstrating strong generalization to more complex real-world scenarios</li>
</ul>
<p>
These results validate the effectiveness of our reinforcement learning approach in enhancing NL2SQL reasoning capabilities, particularly in handling complex queries involving multi-table joins, nested subqueries, and advanced SQL operations.
</p>
<!-- Results table/figure container -->
<div class="columns is-centered has-text-centered experiment-results" style="margin-top: 1.5rem;">
<div class="column">
<img src="images/table1.png" alt="Experiment Results" style="width: 100%; height: auto; border: 1px solid #ddd; border-radius: 6px;"/>
<p class="has-text-grey" style="margin-top: 0.5rem; font-size: 0.9rem;">
<strong>Table 1:</strong> Performance comparison on Spider and BIRD benchmarks. SQL-R1 shows consistent improvements across different model sizes.
</p>
</div>
</div>
</div>
<!-- Model Variants -->
<div class="experiment-subsection" style="margin-top: 2rem;">
<h3 class="title is-4">Model Variants and Scalability</h3>
<p>
We release SQL-R1 in three different sizes to accommodate various deployment scenarios:
</p>
<ul>
<li><strong>SQL-R1-3B:</strong> Lightweight model suitable for resource-constrained environments</li>
<li><strong>SQL-R1-7B:</strong> Balanced performance and efficiency for general use cases</li>
<li><strong>SQL-R1-14B:</strong> Maximum performance for demanding applications</li>
</ul>
<p>
All variants demonstrate consistent improvements over their respective supervised baselines, confirming the scalability of our reinforcement learning approach across different model capacities.
</p>
</div>
<!-- Additional Experimental Details -->
<div class="experiment-subsection" style="margin-top: 3rem;">
<h3 class="title is-4">Additional Experimental Details</h3>
<p style="margin-bottom: 2rem;">
Our paper provides comprehensive experimental analysis beyond the results presented above. For detailed information on the following aspects, we refer readers to the full paper:
</p>
<div style="background: linear-gradient(to right, #f8f9fa 0%, #ffffff 100%); border: 1px solid #e1e4e8; padding: 2.5rem; margin: 2rem 0; border-radius: 8px; box-shadow: 0 2px 8px rgba(0,0,0,0.08);">
<div class="columns is-multiline" style="margin-bottom: 0;">
<div class="column is-6" style="padding-bottom: 0.5rem;">
<div style="display: flex; align-items: start; margin-bottom: 1.5rem;">
<div style="min-width: 8px; height: 8px; background: #3273dc; border-radius: 50%; margin-top: 0.5rem; margin-right: 1rem;"></div>
<div>
<strong style="color: #2c3e50;">Ablation Studies:</strong>
<p style="color: #555; margin-top: 0.3rem; line-height: 1.6;">Detailed analysis of individual component contributions, including reward function design, training strategies, and data engineering techniques</p>
</div>
</div>
<div style="display: flex; align-items: start;">
<div style="min-width: 8px; height: 8px; background: #3273dc; border-radius: 50%; margin-top: 0.5rem; margin-right: 1rem;"></div>
<div>
<strong style="color: #2c3e50;">Error Analysis:</strong>
<p style="color: #555; margin-top: 0.3rem; line-height: 1.6;">In-depth examination of failure cases and challenging query patterns across different complexity levels</p>
</div>
</div>
</div>
<div class="column is-6" style="padding-bottom: 0.5rem;">
<div style="display: flex; align-items: start; margin-bottom: 1.5rem;">
<div style="min-width: 8px; height: 8px; background: #3273dc; border-radius: 50%; margin-top: 0.5rem; margin-right: 1rem;"></div>
<div>
<strong style="color: #2c3e50;">Comparative Evaluation:</strong>
<p style="color: #555; margin-top: 0.3rem; line-height: 1.6;">Extended comparisons with state-of-the-art methods on additional benchmarks and query types</p>
</div>
</div>
<div style="display: flex; align-items: start;">
<div style="min-width: 8px; height: 8px; background: #3273dc; border-radius: 50%; margin-top: 0.5rem; margin-right: 1rem;"></div>
<div>
<strong style="color: #2c3e50;">Application Scenarios:</strong>
<p style="color: #555; margin-top: 0.3rem; line-height: 1.6;">Discussion of practical deployment considerations and use cases in real-world database systems</p>
</div>
</div>
</div>
</div>
</div>
<div class="has-text-centered" style="margin-top: 2.5rem; padding: 2rem; background: #fafbfc; border-radius: 8px;">
<p style="margin-bottom: 1.5rem; color: #586069; font-size: 1.05rem;">
📄 For complete experimental protocols, additional results, and comprehensive discussions
</p>
<a href="https://arxiv.org/pdf/2504.08600.pdf" target="_blank"
class="button is-link is-medium" style="border-radius: 6px; padding-left: 2rem; padding-right: 2rem; box-shadow: 0 4px 12px rgba(50, 115, 220, 0.3); transition: all 0.3s ease;">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Read the Full Paper</span>
</a>
<p style="margin-top: 1.2rem; font-size: 0.9rem; color: #6a737d; font-style: italic;">
Ma et al., "SQL-R1: Training Natural Language to SQL Reasoning Model By Reinforcement Learning", arXiv:2504.08600, 2025
</p>
</div>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!--BibTex citation -->
<section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title is-3">Citation</h2>
<p class="has-text-centered" style="margin-bottom: 1.5rem; color: #555; font-size: 1.05rem; font-weight: 500;">
If you find SQL-R1 useful for your research, please consider citing our paper:
</p>
<div class="bibtex-container">
<button class="copy-button" onclick="copyBibTeX(this)">
<i class="fas fa-copy"></i> Copy
</button>
<pre><code>@article{ma2025sql,
title={SQL-R1: Training Natural Language to SQL Reasoning Model By Reinforcement Learning},
author={Ma, Peixian and Zhuang, Xialie and Xu, Chengjin and Jiang, Xuhui and Chen, Ran and Guo, Jian},
journal={arXiv preprint arXiv:2504.08600},
year={2025}
}</code></pre>
</div>
</div>
</section>
<!--End BibTex citation -->
<script>
function copyBibTeX(button) {
const code = button.nextElementSibling.querySelector('code');
const text = code.textContent;
navigator.clipboard.writeText(text).then(() => {
const originalText = button.innerHTML;
button.innerHTML = '<i class="fas fa-check"></i> Copied!';
button.classList.add('copied');
setTimeout(() => {
button.innerHTML = originalText;
button.classList.remove('copied');
}, 2000);
}).catch(err => {
console.error('Copy failed:', err);
button.innerHTML = '<i class="fas fa-times"></i> Failed';
setTimeout(() => {
button.innerHTML = '<i class="fas fa-copy"></i> Copy';
}, 2000);
});
}
</script>
<footer class="footer">
<div class="container">
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p>
This page was built using the <a href="https://github.com/eliahuhorwitz/Academic-project-page-template" target="_blank">Academic Project Page Template</a> which was adopted from the <a href="https://nerfies.github.io" target="_blank">Nerfies</a> project page.
You are free to borrow the source code of this website, we just ask that you link back to this page in the footer. <br> This website is licensed under a <a rel="license" href="http://creativecommons.org/licenses/by-sa/4.0/" target="_blank">Creative
Commons Attribution-ShareAlike 4.0 International License</a>.
</p>
</div>
</div>
</div>
</div>
</footer>
<!-- Statcounter tracking code -->
<!-- You can add a tracker to track page visits by creating an account at statcounter.com -->
<!-- End of Statcounter Code -->
</body>
</html>