-
Notifications
You must be signed in to change notification settings - Fork 81
Expand file tree
/
Copy pathindex.html
More file actions
5304 lines (5030 loc) · 278 KB
/
index.html
File metadata and controls
5304 lines (5030 loc) · 278 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
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!DOCTYPE html>
<html lang="en">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
<meta
name="viewport"
content="width=device-width, initial-scale=1, shrink-to-fit=no"
/>
<meta name="description" content="BIRD-bench" />
<meta name="author" content="" />
<link rel="icon" href="./logo/bird_fig_main.png" type="image/x-icon" />
<title>BIRD-bench</title>
<!-- Bootstrap core CSS -->
<link href="./style/bootstrap.min.css" rel="stylesheet" />
<!-- Custom fonts for this template -->
<link
href="./style/font-awesome.min.css"
rel="stylesheet"
type="text/css"
/>
<link
rel="stylesheet"
href="./style/all.css"
integrity="sha384-mzrmE5qonljUremFsqc01SB46JvROS7bZs3IO2EmfFsd15uHvIt+Y8vEf7N7fWAU"
crossorigin="anonymous"
/>
<!-- Custom styles for this template -->
<link href="./style/clean-blog.min.css" rel="stylesheet" />
<link href="./style/bird.css" rel="stylesheet" />
</head>
<body>
<header class="masthead" id="header-color">
<div style="background-color: #485159; text-align: left">
<img
style="margin-top: 5px; padding: 15px"
height="70px"
src="./logo/damo.png"
/>
<img
style="margin-top: 5px; padding: 15px"
height="70px"
src="./logo/hku.png"
/>
<img
style="margin-top: 5px; padding: 15px"
height="70px"
src="./logo/uiuc.png"
/>
</div>
<div class="container">
<div class="row">
<div class="col-lg-10 col-md-12 mx-auto">
<div class="row site-heading flex-box">
<div class="col-lg-3">
<img id="logo-img" src="logo/bird_fig_main.png" alt="logo" />
</div>
<div class="col-lg-9">
<div class="align-middle">
<h1 id="bird-title">BIRD-SQL</h1>
<span class="subheading" id="caption">A Big Bench for Large-Scale Database Grounded
Text-to-SQLs</span>
</div>
</div>
</div>
</div>
</div>
</div>
</header>
<!-- Main Content -->
<div class="container" id="main">
<div class="row">
<div class="col-lg-5">
<div class="list-group">
<div class="list-group-item" id="introduction">
<!-- <div style="display: grid; grid-template-columns: auto auto; align-items: center;"> -->
<h4>About BIRD</h4>
<!-- </div> -->
<p>
BIRD (<b>BI</b>g Bench for La<b>R</b>ge-scale <b>D</b>atabase
Grounded Text-to-SQL Evaluation) represents a pioneering,
cross-domain dataset that examines the impact of extensive
database contents on text-to-SQL parsing.
<!-- The dataset includes three extra problems that especially resonates with real-world scenarios:
<b>1)</b> the necessity to incorporate external knowledge into parsers; <b>2)</b> the need for parsers to
reason over 'dirty' values;
and <b>3)</b> the requirment to consider efficiency while executing SQLs. -->
BIRD contains over <b>12,751</b> unique question-SQL pairs,
<b>95</b> big databases with a total size of <b>33.4 GB</b>. It
also covers more than <b>37</b> professional domains, such as
blockchain, hockey, healthcare and education, etc.
</p>
<div id="button-group">
<div class="button-row">
<a
href="https://livesqlbench.ai/data-intelligence-index/"
target="_blank"
>
<button type="button" class="btn btn-g">
🔥 Data Intelligence Index
</button>
</div>
<div class="button-row">
<a
href="https://arxiv.org/pdf/2305.03111.pdf"
target="_blank"
>
<button type="button" class="btn btn-a">Paper</button>
</a>
</div>
<div class="button-row">
<a
href="https://github.com/AlibabaResearch/DAMO-ConvAI/tree/main/bird"
target="_blank"
>
<button type="button" class="btn btn-a">Code</button>
</a>
</div>
<div class="button-row">
<a
href="https://github.com/bird-bench/mini_dev"
target="_blank"
>
<button type="button" class="btn btn-a">
Mini-Dev (500)
</button>
</a>
</div>
<div class="button-row">
<a
href="https://bird-critic.github.io/"
target="_blank"
>
<button type="button" class="btn btn-d">
BIRD-CRITIC 1.0 (SQL)
</button>
</div>
<div class="button-row">
<a
href="https://livesqlbench.ai/"
target="_blank"
>
<button type="button" class="btn btn-e">
LiveSQLBench!
</button>
</div>
<div class="button-row">
<a
href="https://bird-interact.github.io/"
target="_blank"
>
<button type="button" class="btn btn-d">
BIRD-Interact
</button>
</div>
<div class="button-row" id="button-row-b">
<div class="btn-round">
<a
href="https://bird-bench.oss-cn-beijing.aliyuncs.com/train.zip"
target="_blank"
>
<button type="button" class="btn btn-b">Train Set</button>
</a>
</div>
<div class="btn-round">
<a
href="https://bird-bench.oss-cn-beijing.aliyuncs.com/dev.zip"
target="_blank"
>
<button type="button" class="btn btn-b">
🔥 Dev Set
</button>
</a>
</div>
</div>
</div>
</div>
<style>
.scroll-container {
position: relative;
width: 100%;
max-height: 300px;
overflow-y: scroll;
}
/* Hide the scrollbar by default */
.scroll-container::-webkit-scrollbar {
width: 10px;
background: transparent;
}
/* Set the scrollbar thumb color */
.scroll-container::-webkit-scrollbar-thumb {
background-color: #ffffff;
border-radius: 5px;
}
/* Show the scrollbar when the mouse is hovering over the scroll container */
.scroll-container:hover::-webkit-scrollbar {
width: 10px;
}
.scroll-container:hover::-webkit-scrollbar-thumb:hover {
background-color: #555;
}
</style>
<script>
// A new function for BIRD-CRITIC
function showBirdCriticMessage() {
alert(
"BIRD-CRITIC is the first SQL debugging benchmark investigating whether LLMs can effectively resolve user issues in real-world database applications. It contains 600 tasks in dev and 200 OOD hidden tests. The full challenges in BIRD-SQL 2025 are under collaboration with Google Cloud.\n\n" +
"Key Highlights:\n" +
"• 5 popular SQL dialects: MySQL, PostgreSQL, SQLServer, Oracle, and one OOD dialect\n" +
"• Challenging tasks that go beyond SELECT (e.g., CREATE TRIGGER, UPDATE constraints, SQL Optimization)\n" +
"• Each task is derived from real-world user cases and reproduced within the BIRD environment, verified by human experts\n" +
"• Multi-thread sandbox enabling fast and robust evaluation\n" +
"• Professionally curated test cases for each task.\n" +
"We will release a PREVIEW set in this month and full set next month. Email us if you have some suggestions or new contents that you want to see!\n\n" +
"Other exciting Sibling benchmarks you may expect in BIRD-SQL 2025:\n" +
"• BIRD-CRITIC 1.5 (Repo-level Code Debugging for Real-World DB Applications)\n" +
"• BIRD-Interact\n" +
"• BIRD-Vision\n" +
"• BIRD-Pro & BIRD-Pro Max: The second generation of BIRD-SQL covering comprehensive DB operations / BI querying with new open-source enterprise-level databases with thousands of columns, advanced constrains and TB-level data."
);
}
</script>
<script>
function showLiveSQLBenchMessage() {
alert(
"Inspired by great LiveCodeBench, we're developing LiveSQLBench to provide a robust and contamination-free evaluation for text-to-SQL models on challenging, comprehensive benchmark, which will be updated periodically.\n\n" +
"LiveSQLBench (BIRD-SQL Pro v0.5) will feature diverse real-world user queries, including Business Intelligence (BI), CRUD operations, etc.\n\n" +
"Each release will include 50 new, fully open-source DBs curated by the BIRD team through expert collaboration and continuous improvement. It will cover a wide range of database sizes, from end-user level (around 39 columns) to industrial level (1340+ columns), catering to various query types.\n\n" +
"Look for LiveSQLBench to launch in March! And more exciting text-to-SQL evaluations on the way~ :) "
);
}
</script>
<div class="list-group-item" id="news">
<h4>News</h4>
<div class="scroll-container">
<div
class="card card-outline-secondary mb-4"
style="text-align: left"
>
<div class="card-body" style="background-color: #f1f6f9">
<ul style="padding-left: 0">
<li style="list-style-type: none">
<span style="background-color: #634687; color: white; padding: 2px 4px; border-radius: 5px;">
<strong style="color: #ffe7ce">🇧🇷 Apr. 22, 2026: </strong> </span> Arrived in Rio 🇧🇷, and we are looking forward to oral presenting our ICLR 2026 work: <strong><a href="https://bird-interact.github.io/" target="_blank">BIRD-Interact</a></strong>. Feel free to contact us if you’d like to chat 🗓 Oral: Fri, Apr 24, 2026 • 3:27–3:37 PM (Room 203 A/B); 🪧 Poster: Poster Session 3, Pavilion 3, 10:30 AM – 1:00 PM
</li>
<li style="list-style-type: none">
<span style="background-color: #B03052; color: white; padding: 2px 4px; border-radius: 5px;">
<strong style="color: #ffe7ce">🚀 Mar. 23, 2026:</strong>
</span>
We release <strong><a href="https://huggingface.co/datasets/birdsql/bird-critic-1.0-sqlite" target="_blank" style="color:#155E95; text-decoration: none;">BIRD-Critic-SQLite</a></strong>, a new single-dialect benchmark containing <strong>500</strong> high-quality user issues focused on real-world SQLite database applications. Along with the dataset, we also release 3 RL-trained models: <strong><a href="https://huggingface.co/birdsql/BIRD-Talon-14b" target="_blank" style="color:#155E95; text-decoration: none;">BIRD-Talon-14B</a></strong>, <strong><a href="https://huggingface.co/birdsql/BIRD-Talon-7b" target="_blank" style="color:#155E95; text-decoration: none;">BIRD-Talon-7B</a></strong>, and <strong><a href="https://huggingface.co/birdsql/BIRD-Zeno-7b" target="_blank" style="color:#155E95; text-decoration: none;">BIRD-Zeno-7B</a></strong>. Data is available on <strong><a href="https://huggingface.co/datasets/birdsql/bird-critic-1.0-sqlite" target="_blank" style="color:#155E95; text-decoration: none;">HuggingFace</a></strong> and evaluation script on <strong><a href="https://github.com/bird-bench/BIRD-CRITIC-1" target="_blank" style="color:#155E95; text-decoration: none;">GitHub</a></strong>.
</li>
<li style="list-style-type: none">
<span style="background-color: #B03052; color: white; padding: 2px 4px; border-radius: 5px;">
<strong style="color: #ffe7ce">Mar. 3, 2026: </strong> </span>
<strong><a href="https://huggingface.co/datasets/birdsql/livesqlbench-large-v1" target="_blank">LiveSQLBench-Large-v1</a></strong> is released!
The industrial-scale counterpart featuring ~1K columns, ~54 tables each DB with 480 tasks,
10x schema complexity over LiveSQLBench-Base-Full-v1, ~84K avg prompt tokens for long-context challenge,
and Business Rule Drift (external knowledge can change and even has inconsistency across releases) for live context-learning evaluation.
See more at <strong><a href="https://livesqlbench.ai" target="_blank">livesqlbench.ai</a></strong>.
</li>
<li style="list-style-type: none">
<span style="background-color: #B03052; color: white; padding: 2px 4px; border-radius: 5px;">
<strong style="color: #ffe7ce">Mar. 6, 2026: </strong> </span>
We release the <strong><a href="https://livesqlbench.ai/data-intelligence-index/" target="_blank">Data Intelligence Index</a></strong>, a comprehensive evaluation of frontier AI models and agents on data-centric intelligence, covering DB querying, BI analysis, DB application debugging, human-centric interaction, multi-modal querying, and more.
<strong>If you'd like your benchmark or model to be included, please feel free to <a href="mailto:bird.bench25@gmail.com">contact us</a>!</strong>
</li>
<li style="list-style-type: none">
<span style="background-color: #634687; color: white; padding: 2px 4px; border-radius: 5px;">
<strong style="color: #ffe7ce">🇧🇷 Feb. 8, 2026: </strong> </span> Our paper <strong><a href="https://arxiv.org/abs/2510.05318" target="_blank">BIRD-Interact</a></strong> has been accepted to <strong>ICLR 2026 (Oral)</strong>!
See you in <strong>Rio 🇧🇷</strong>！
</li>
<li style="list-style-type: none">
<span style="background-color: #634687; color: white; padding: 2px 4px; border-radius: 5px;">
<strong style="color: #ffe7ce">🌴 Nov. 30, 2025: </strong> </span> Our paper <strong><a href="https://arxiv.org/abs/2506.18951" target="_blank">SWE-SQL</a></strong> has been accepted to <strong>NeurIPS 2025 (main track)</strong>!
You can find us in <strong>San Diego 🌴 🦞</strong>.
Feel free to contact us if you’d like to chat:
<a href="mailto:jl0725@connect.hku.hk">Patrick</a>,
<a href="mailto:xia01ong@connect.hku.hk">Xiaolong Li</a>.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #B03052;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Nov. 13, 2025:</strong></span>
<span
style="
background-color: #FF6B35;
color: white;
padding: 2px 6px;
border-radius: 5px;
margin-left: 5px;
font-weight: bold;
"
>Major Updates</span>
<br><br>
<strong>1.</strong> Our team with the efforts of global engineers, experts, and students has completed comprehensive quality control for BIRD-SQL, releasing <a href="https://huggingface.co/datasets/birdsql/bird_sql_dev_20251106" target="_blank"><strong>bird-sql-dev-1106</strong></a>, which is a cleaner development split. For new submissions using this updated split, please indicate this in your submission so we can mark it accordingly on the leaderboard.
To better resolve ambiguity, we will open a new track of interactive setting as we did in BIRD-Interact and remove Oracle Evidence instead. Coming soon!
<br><br>
<strong>2.</strong> We have released <a href="https://huggingface.co/datasets/birdsql/mini-interact" target="_blank"><strong>Mini-Interact</strong></a>, an SQLite-based interactive benchmark with a self-contained environment that requires no Docker setup. This lightweight version enables fast development and testing of interactive intelligence for agentic data exploration with users.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #B03052;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Oct. 9, 2025:</strong></span> We release our paper of <a href="https://huggingface.co/papers/2510.05318" target="_blank"><strong>BIRD-Interact</strong></a>, detailing how we built the dataset and our key findings on why interaction and communication are critical for frontier LLMs to become more reliable assistance in the DBA cycle.
Currently, GPT-5 (Med) only achieves 8.67% SR on <code>c-Interact</code> and 17.00% SR on <code>a-Interact</code> of full tasks. Additionally, we will release <code>Mini-Interact</code> which will be operated in sqlite, with our trained local and stable user simulator. It is currently undergoing alignment and function testing before release.
If you think our work is helpful to you, we will feel grateful if you can star us, which is a strong motivation for boosting our open-source efforts!
</li>
<li style="list-style-type: none">
<span
style="
background-color: #B03052;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Sep. 19, 2025:</strong></span> We’ve released <a href="https://huggingface.co/datasets/birdsql/bird23-train-filtered" target="_blank"><strong>BIRD23-train-filtered</strong></a>, a high-quality filtered subset of the BIRD train split (6,601/9,428 ≈70%) that works as a drop-in replacement for text-to-SQL finetuning; See the example finetuning usage in <a href="https://github.com/bird-bench/mini_dev/tree/main/finetuning" target="_blank">here</a>.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #B03052;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Sep. 10, 2025:</strong></span> We're excited to announce the release of the <a href="https://huggingface.co/datasets/birdsql/livesqlbench-base-full-v1" target="_blank"><strong>LiveSQLBench-Base-Full-V1 (600)</strong></a>! The first text-to-SQL benchmark covering all SQL spectrum with Hierarchical Knowlegde Base (HKB) and test cases.
We provide two types of queries: normal query and colloquial queries for people to test according to their own needs. The flag model Gemini-2.5-pro can only achieve <strong>28.67</strong> in colloquial queries, and <strong>35.67</strong> in normal queries. The <a href="https://huggingface.co/datasets/birdsql/livesqlbench-base-lite" target="_blank">base-lite</a> and <a href="https://huggingface.co/datasets/birdsql/livesqlbench-base-full-v1" target="_blank">base-full-v1</a> would be locked version for development of research methods.
The detailed performance is in our <a href="https://livesqlbench.ai/" target="_blank">website</a>.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #B03052;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Aug. 26, 2025:</strong></span> We're excited to announce the release of the <a href="https://huggingface.co/datasets/birdsql/bird-interact-full" target="_blank"><strong>BIRD-Interact-Full (600)</strong></a> set! It's a tough one—the best LLMs are only achieving a <strong>16.33%</strong> success rate, with just <strong>10.0%</strong> on the <code>c-interact</code> and <code>a-interact</code> portions. For more details, please visit our <a href="https://bird-interact.github.io/" target="_blank">project website</a>.
We'll be sending the Ground Truth & Test cases to our mailing list this week. If you want early access, please send an email as instructed on the site for an automatic download. On another note, we've also released a SQLite version of <strong><a href= "https://huggingface.co/datasets/birdsql/livesqlbench-base-lite-sqlite">LiveSQLBench-Lite</a></strong> for easier local research. The full <strong>LiveSQLBench-Base</strong> and <strong>-Large</strong> versions are coming soon!
</li>
<li style="list-style-type: none">
<span
style="
background-color: #68488f;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Jul. 15, 2025:</strong></span> LLMs always struggle with self-correction due to self-enhancement bias, a problem particularly pronounced in SQL generation where declarative syntax and concise logs would provide limited guidance. Our new work in <code>ACL Main</code>, <strong>SHARE</strong>, mitigates this significantly by (1) converting SQL queries into procedural programming steps and (2) proposing a generalized pipeline through on-policy multi-agent methods for fine-tuning SLMs.
<br><br>
<strong>Models:</strong>
<a href="https://huggingface.co/birdsql/share-bam">BAM</a>,
<a href="https://huggingface.co/birdsql/share-sam">SAM</a>,
<a href="https://huggingface.co/birdsql/share-lom">LOM</a>
<br>
<strong>Dataset:</strong> <a href="https://huggingface.co/datasets/birdsql/share-bam">https://huggingface.co/datasets/birdsql/share-bam</a>
<br>
<strong>Paper:</strong> <a href="https://huggingface.co/papers/2506.00391">https://huggingface.co/papers/2506.00391</a>
<br>
<strong>Code:</strong> <a href="https://github.com/quge2023/SHARE">https://github.com/quge2023/SHARE</a>
</li>
<li style="list-style-type: none">
<span
style="
background-color: #68488f;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Jul. 10, 2025:</strong></span> We optimized and upload <a href="https://huggingface.co/datasets/birdsql/bird_mini_dev"><strong>bird-mini-dev</strong></a> with 3 dialects to huggingface according to community. Thanks!! Please download and check!
</li>
<li style="list-style-type: none">
<span
style="
background-color: #68488f;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Jul. 10, 2025:</strong></span> We release the human performance scores on <a href="https://bird-critic.github.io/"><strong>BIRD-CRITIC 1.0</strong></a>! It contains both the group with and without usage of GenAI Tools. The huge gap encourages more intelligent solutions to serve users in the real-world DB applications. Please check~~.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #68488f;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Jun. 25, 2025:</strong></span> We document our procedures, findings, and training recipe for open-source LLM agents that resolve SQL debugging tasks in our paper <a href="https://huggingface.co/papers/2506.18951">SWE-SQL</a>.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #68488f;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Jun. 08, 2025:</strong></span> We have released bird-critic-1.0-postgresql (sql issues for single dialects). Check out the data in <a href="https://huggingface.co/datasets/birdsql/bird-critic-1.0-postgresql">Hugging Face</a> and the newest code in <a href="https://github.com/bird-bench/BIRD-CRITIC-1">GitHub</a>.
It seems that bird-critic is a challenging reasoning tasks for text-to-SQL since all top-peforming models are reasoning-based models. Have fun! Thanks!
</li>
<li style="list-style-type: none">
<span
style="
background-color: #68488f;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Jun 4, 2025:</strong></span> We release <a href="https://bird-interact.github.io/">BIRD-Interact</a>, a comprehensive interactive evaluation for text-to-SQL models. It contains conversational (<code>c-Interact</code>) and agentic (<code>a-Interact</code>) interaction modes.
Top results: o3-mini achieves <strong>24.4%</strong> Success Rate on <code>c-Interact</code>, while Claude-3.7-Sonnet reaches <strong>17.78%</strong> Success Rate on <code>a-Interact</code>. We also discovered <strong>Interaction-Time Scaling (ITS)</strong> - performance scales over extended interactions. Enjoy!
</li>
<li style="list-style-type: none">
<span
style="
background-color: #68488f;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">May 29, 2025:</strong></span> We release <a href="https://livesqlbench.ai/">LiveSQLBench</a>, the first containmination-free text-to-SQL benchmark covering full SQL specturm featuring more advanced SQLs, annotated RDBs, structrued / unstructured knowledge base, test cases, etc.
First, we present <strong><code>LiveSQLBench-Base-Lite</code></strong> with 270 tasks for trail. Even though Lite is the easiest version with the most clear conditions, the SOTA model <code>o3-mini</code> only achieves <strong>44.81%</strong> Success Rate.
We will update full set and settings. You can preview and interact with data samples in our website. Stay tuned!
</li>
<li style="list-style-type: none">
<span
style="
background-color: #22a699;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce"
>May. 25, 2025:</strong></span> Please report issues when you find any more of BIRD-SQL 2023,
we will check and clean dev set (1534) for the last time during summer by considering all feedbacks.
Next week, we will start our new project <strong><code>LiveSQLBench</code></strong>, the first one-stop containmination-free text-to-SQL benchmark covering full SQL specturm featuring more advanced SQLs,
databases, hierarichal / unstructured knoweldge base, test cases.
Also it supports multi-turn conversational and interactive evaluation via our <strong><code>BIRD-Interact</code></strong>, which will be released together.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #B03052;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">May 22, 2025:</strong></span> We update Single-Model Leaderboard, now the Self-Consistency column has 4 values:
<strong>empty</strong> (no self consistency used);
<strong>Few:</strong> 1-7 candidates used in majority vote;
<strong>Many:</strong> 8-32 candidates;
<strong>Scale:</strong> >32 candidates;
</li>
<li style="list-style-type: none">
<span
style="
background-color: #B03052;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Apr. 20, 2025:</strong></span> We have released bird-critic-1.0-open (600 tasks by 4 dialects). Check out the data in <a href="https://huggingface.co/datasets/birdsql/bird-critic-1.0-open">Hugging Face</a> and the newest code in <a href="https://github.com/bird-bench/BIRD-CRITIC-1">GitHub</a>. The full set of PostgreSQL will be released 1 week later.
It seems that bird-critic is a challenging reasoning tasks for text-to-SQL since all top-peforming models are reasoning-based models. Have fun! Thanks!
</li>
<li style="list-style-type: none">
<span
style="
background-color: #B03052;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Feb. 4, 2025:</strong></span> We've launched <a href="https://bird-critic.github.io/">BIRD-Critic (a.k.a SWE-SQL)</a>, a brand new text-to-SQL benchmark that really digs into reasoning challenges! A lite version is ready for exploration. Full sets are coming soon! Feel free for any feedbacks! Your inputs and suggestions are much appreciated!
</li>
<!-- <li style="list-style-type: none;"><span style="background-color: #22A699; color: white; padding: 2px 4px; border-radius: 5px;"><strong style="color: #FFE7CE">May. 27, 2024:</strong></span> Glad to share our paper <a href="https://arxiv.org/abs/2405.15307">Before Generation, Align it! A Novel and Effective Strategy for Mitigating Hallucinations in Text-to-SQL Generation (TA-SQL)</a>, which will appear at <code><strong>ACL 2024</strong></code> Findings. This is a light but effective method to calbriate hallucinations of text-to-SQL generation. Additionally, we firstly introduce Pythonic prompts for text-to-SQL and include results on open-source LLMs.</li> -->
<li style="list-style-type: none">
<span
style="
background-color: #B03052;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce">Nov. 26, 2024:</strong></span> Thanks the support of <code>BIRD-SQL 2023</code>!
Now we are pleased to share that the project <strong><code>BIRD 2025</code></strong> has been started.
It will contains 4-6 new benchmarks with each covering its special focus of professional databases and their knowledge in the wild applications.
<strong>We will release the first benchmark by early Jan</strong>. Feel free to let us know your needs or suggestions for cooking new generations of Text-to-SQL challenges. Thanks!</a>
</li>
<li style="list-style-type: none">
<span
style="
background-color: #22a699;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce"
>Aug. 4, 2024:</strong></span> The Reward-based Valid Efficiency Score (R-VES) will be used as the efficiency metric for future test submissions.
The rationale and formula for R-VES can be found in the <a href="https://github.com/bird-bench/mini_dev">Mini-Dev repository</a>.
You can check the legacy VES scores for previous submissions <a href="https://abundant-dawn-033.notion.site/BIRD-VES-Leaderboard-Legacy-7aea622573dd42c59a1bcdb122d2cf44?pvs=4">here</a>.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #4f709c;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce"
>Jun. 30, 2024:</strong
></span
> Due to large requests of test submissions about mixed
models (open-source + GPU-based closed source), we
update the submission instructions to accelerate your
waiting time. Please check it out!
</li>
<li style="list-style-type: none">
<span
style="
background-color: #4f709c;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce"
>Jun. 30, 2024:</strong
></span
>
If you are interested in code agent, please do not miss a SOTA code agent implementation by <a href="https://github.com/All-Hands-AI/OpenHands/tree/main/evaluation/bird">OpenDevin</a> for BIRD Dev!
</li>
<li style="list-style-type: none">
<span
style="
background-color: #4f709c;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce"
>Jun. 27, 2024:</strong
></span
>
Excited to announce the release of our
<a
ref="https://github.com/bird-bench/mini_dev/tree/main"
>BIRD Mini-Dev dataset</a
>
with 500 high-quality examples. This dataset includes
all BIRD keywords, with modifications for questions such
as the addition of <code>window function</code>. We are
the first to deliver it in not only
<strong>SQLite</strong>, but also
<strong>MySQL</strong>, and <strong>PostgreSQL</strong>.
We include Soft-F1 and R-VES metrics to reduce bias.
Don't miss the <code>column_meaning.json</code> file,
preprocessed by TA-SQL. Available for dev and testing
set. Check out our work here:
<a href="https://arxiv.org/abs/2405.15307"
>Before Generation, Align it! A Novel and Effective
Strategy for Mitigating Hallucinations in Text-to-SQL
Generation (TA-SQL)</a
>, appearing at
<code><strong>ACL 2024</strong></code> Findings.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #4f709c;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce"
>Apr. 27, 2024:</strong
></span
>
Due to large volume of requests, we now modify the
license of our data to
<code><strong>CC BY-SA 4.0</strong></code
>. However, we will not take responsibility of any bad
purposes by using our data. Since we develop this
benchmark for research and healthy application only.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #4f709c;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce"
>Mar 13, 2024:</strong
></span
>
Please also take a look at our related work:
<a href="https://arxiv.org/abs/2403.05307"
>Tapilot-Crossing</a
>, which is the first challenging and more realistic
benchmark designed to evaluate Large Language Model
(LLM) agents on interactive data analysis tasks. The
code includes Python and Private Library. And it covers
6 common agent actions in evaluation.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #4f709c;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce"
>Sept 25, 2023:</strong
></span
>
We have released a cleaner version of
<code>dev set</code>. Please download dev set again. We
checked all cases of dev set and fixed all errors that
we found. After cleaning, the ChatGPT (gpt-3.5-turbo)
and GPT4 (gpt-4-32k) EX scores have improved to
<strong>42.24</strong> (from 37.22) and
<strong>49.15</strong> (from 46.35), respectively.
Thanks for all feedbacks!
</li>
<li style="list-style-type: none">
<span
style="
background-color: #4f709c;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #ffe7ce"
>Sept 21, 2023:</strong
></span
>
Our paper has been accepted by
<code><strong>NeurIPS 2023</strong></code> as a
<code><strong>Spotlight</strong></code
>!!! Thanks for all the efforts and suggestions of
co-authors, anonymous reviewers, awesome
researchers/users in github or emails.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #4f709c;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #f6f1f1"
>July 17, 2023:</strong
></span
>
We update newest results of GPT-4, Claude-2 and Palm-2.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #4f709c;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #f6f1f1"
>July 14, 2023:</strong
></span
>
The data link has been updated, fixing the schema names
in the CSV files. Additionally, tied results caused by
<code>order_by limit 1</code> are now considered. Both
SQL queries - with and without accounting for tied
results - are valid at this time.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #4f709c;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #f6f1f1"
>Jun 12, 2023:</strong
></span
>
We are welcome to any suggestions and reported gold
errors in
<a
href="https://github.com/AlibabaResearch/DAMO-ConvAI/issues/39"
>help_wanted</a
>. Any of your help is appreciated!
</li>
<li style="list-style-type: none">
<span
style="
background-color: #4f709c;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #f6f1f1"
>Jun 5, 2023:</strong
></span
>
We open-sourced our
<a href="https://arxiv.org/abs/2301.07507">Graphix-T5</a
>, a graph-aware semi-pretrained text-to-text PLM
specifically designed to improve multi-hop reasoning for
the complex text-to-SQL task.
</li>
<li style="list-style-type: none">
<span
style="
background-color: #4f709c;
color: white;
padding: 2px 4px;
border-radius: 5px;
"
><strong style="color: #f6f1f1"
>May 30, 2023:</strong
></span
>
If you are interested in ICL, please check out our
interesting work
<a href="https://arxiv.org/pdf/2305.13016.pdf"
>deep-thinking🤔.</a
>
Generate 1000 models for 1000 people smoothly!
</li>
</ul>
</div>
</div>
</div>
</div>
<div class="list-group-item" id="illustration">
<h4>Surprise from BIRD</h4>
<p>
<b>1. Large and Dirty values:</b> Due to the nature of the
real-world scenarios from which BIRD's database values were
collected, they typically retain their original and frequently
"dirty" format. Hence, text-to-SQL parsers must first analyze
these values to account for their non-standard format before
engaging in reasoning.
</p>
<img
src="./img/dirty_1.png"
class="column-img"
id="dirty-img-1"
style="margin: 0 auto"
/>
<img
src="./img/dirty_2.png"
class="column-img"
id="dirty-img-2"
style="margin: 0 auto"
/>
<p>
<b>2. External Knowledge:</b> "account.type = 'OWNER'" can be
inferred by the knowledge evidence: "The condition of the loans
require the account type should be the owner."
</p>
<img
src="./img/ex_kg_2.png"
class="column-img"
id="kg-img"
style="margin: 0 auto"
/>
<p>
<b>3. Text-to-Efficient-SQL:</b> BIRD is the first text-to-SQL
benchmark designed to encourage semantic parsers to produce SQL
queries that are not only correct but also efficient. This
emphasis on efficiency is especially valuable in real-world data
/ business analysis circumstances.
</p>
<img
src="./img/efficient_sql.png"
class="column-img"
id="eff-img"
style="margin: 0 auto"
/>
</div>
<!-- <div class="list-group-item">
<h4>Download Dataset</h4>
<p>We are cleaning and checking our databases, SQLs, Question, and Knowledge carefully. Will come up soon!</p>
<p class="text-center">
<a href="https://bird-bench.github.io/"><button type="button" class="btn btn-success", id="code">BIRD Code Nest</button></a>
</p>
</div> -->
<div class="list-group-item">
<h4>Submission</h4>
<p>
Please follow the Submission Guideline (below) and contact
<code>bird.bench23@gmail.com</code> for test evaluation.
Ususally, we will return your results in 10 days!
</p>
<style>
.btn-a {
border-radius: 25px; /* Adjust this value to change the roundness of the button */
}
.btn-a:hover {
background-color: #22a699; /* Change the background color when the cursor is over the button */
}
</style>
<div class="button-row">
<a
href="https://docs.google.com/document/d/1Rs6d_pcs2vfqW4Ymub7Wb1XtBNlrc-WfH3T7U1ktuBo/edit?usp=sharing"
target="_blank"
>
<button type="button" class="btn btn-a">
Submission Guidelines
</button>
</a>
</div>
</div>
<div class="list-group-item">
<h4>Subscribe to BIRD Update</h4>
<p>
Bird is a long-term research project aimed at bridging the gap
between semantic parsing models and the success of database
applications. To receive the latest updates of the dataset, you
can leave your email address.
</p>
<form onsubmit="return submitForm();">
<input
type="email"
id="email"
name="email"
placeholder="Email address"
required
/>
<button type="submit">Subscribe</button>
</form>
<p id="message"></p>
</div>
<head>
<meta charset="utf-8" />
<title>Email Subscription</title>
<script>
function submitForm() {
var email = document.getElementById("email").value;
var url =
"https://script.google.com/macros/s/AKfycbynv6Ibnl9gJosprkRmXhbKtllZX8dqrbj_xQkcbZ_NgEDBxwyjz1xyvIvnm_4yPh3ZPA/exec";
var data = "email=" + encodeURIComponent(email);
fetch(url, {
method: "POST",
body: data,
headers: {
"Content-Type": "application/x-www-form-urlencoded",
},
})
.then((response) => response.text())
.then((text) => {
document.getElementById("message").innerHTML =
"<i>Subscribe Successfully! Thanks for your attention!</i>";
})
.catch((error) => {
console.error("Error:", error);
});
return false;
}
</script>
</head>
<div class="list-group-item">
<h4>Citation</h4>
<a href="https://visitorbadge.io/status?path=https%3A%2F%2Fbird-critic.github.io%2F"><img src="https://api.visitorbadge.io/api/visitors?path=https%3A%2F%2Fbird-critic.github.io%2F&label=Views&labelColor=%23155e95&countColor=%23e9ecef&style=flat" /></a>
<!-- <pre id="citation">
@misc{li2023llm,
title={Can LLM Already Serve as A Database Interface? A BIg Bench for Large-Scale Database Grounded Text-to-SQLs},
author={Jinyang Li and Binyuan Hui and Ge Qu and Binhua Li and Jiaxi Yang and Bowen Li and Bailin Wang and Bowen Qin and Rongyu Cao and Ruiying Geng and Nan Huo and Chenhao Ma and Kevin C. C. Chang and Fei Huang and Reynold Cheng and Yongbin Li},
year={2023},
eprint={2305.03111},
archivePrefix={arXiv},
primaryClass={cs.CL}
}</pre> -->
<pre id="citation">
@article{li2024can,
title={Can llm already serve as a database interface? a big bench for large-scale database grounded text-to-sqls},
author={Li, Jinyang and Hui, Binyuan and Qu, Ge and Yang, Jiaxi and Li, Binhua and Li, Bowen and Wang, Bailin and Qin, Bowen and Geng, Ruiying and Huo, Nan and others},
journal={Advances in Neural Information Processing Systems},
volume={36},
year={2024}
}</pre
>
</div>
</div>
</div>
<div class="col-lg-7">
<div class="btn-group mb-3 w-100">
<button id="overall-leaderboard-btn" class="btn btn-f">Overall Leaderboard</button>
<button id="single-model-leaderboard-btn" class="btn btn-f">Single-Model Leaderboard</button>
</div>
<div id="overall-leaderboard">
<div class="card card-outline-secondary">
<div class="card-header">Leaderboard - Execution Accuracy (EX)</div>
<div class="table-wrapper">
<table class="table" id="ex-table">
<thead class="thead-light">
<tr>
<th scope="col" rowspan="1"></th>
<th
scope="col"
rowspan="1"
class="align-middle text-center"
>
Model
</th>
<th
scope="col"
rowspan="1"
class="align-middle text-center"
>
Code
</th>
<th
scope="col"
rowspan="1"
class="align-middle text-center"
>
Size
</th>
<th
scope="col"
rowspan="1"
class="align-middle text-center"
>
Oracle Knowledge
</th>
<th scope="col" colspan="1" class="text-center">Dev (%)</th>
<th scope="col" colspan="1" class="text-center">
Test (%)
</th>
</tr>
</thead>
<tbody>
<tr class="table-success">
<td scope="row" class="align-middle text-center"><br /></td>
<td class="align-middle text-center">
Human Performance<br />
<span class="affiliation"
>Data Engineers + DB Students</span
><br /><a href="" class="paperlink"></a>
</td>
<td class="align-middle text-center"></td>
<td class="align-middle text-center"></td>
<td class="align-middle text-center">✔️</td>
<td class="align-middle text-center"></td>
<td class="align-middle text-center"><b>92.96</b></td>
</tr>
<!-- <tr class="table-success">
<td scope="row" class="align-middle text-center"><br>
</td>
<td class="align-middle text-center">Human Performance<br>
<span class="affiliation">Data Engineers + DB Students</span><br><a href="" class="paperlink"></a>
</td>
<td class="align-middle text-center"></td>
<td class="align-middle text-center">✔️</td>
<td class="align-middle text-center"></td>
<td class="align-middle text-center"></td>
<td class="align-middle text-center"><b>72.37</b></td>
</tr> -->