List your company

ResearchFormerly WizWand

CoSQL

Benchmarks

TaskDatasetSOTA resultTrendResults

Context-dependent Text-to-SQL

CoSQL (dev)

58.8Question Match

33

Multi-turn Text-to-SQL

CoSQL In-domain

65.2Exact Match

29

Multi-turn Text-to-SQL

CoSQL (dev)

63.98Query-level EM

16

Multi-turn Text-to-SQL

CoSQL Out-of-domain

74Execution Accuracy

16

Conversational SQL

CoSQL

65.83Accuracy

14

CoSQL

-0.5Kendall's Tau

12

Context-dependent Text-to-SQL

CoSQL (test)

55.7Question Match

12

Continuous improvement of language agents

CoSQL StreamBench (32 held-out episodes)

73.5Pass@4

5

Label-free Performance Estimation

CoSQL

7.9MAE (ATHENA)

5

Conversational Semantic Parsing

CoSQL (dev)

56.9Match

5

Dialog state tracking

CoSQL (test)

54.6QM %

3

Dialog state tracking

CoSQL (dev)

0.569QM%

3

Language-to-SQL Parsing

CoSQL (test)

40.8EM

3

Language-to-SQL Parsing

CoSQL (dev)

42EM

3

Multi-turn Text-to-SQL

CoSQL ADVETA-ADD

32.8Exact Match (EM)

2

Multi-turn Text-to-SQL

CoSQL (ADVETA-RPL)

16.4Exact Match (EM)

2

Multi-turn Text-to-SQL

CoSQL original (dev)

44.1Exact Match (EM)

2