Dataset Viewer
Auto-converted to Parquet Duplicate
qid
stringlengths
5
5
task_type
stringclasses
1 value
has_intended_resolution
bool
1 class
db
stringclasses
6 values
question
stringlengths
33
189
gold_ambiguity_points
listlengths
1
5
gold_queries
listlengths
1
32
gold_intended_query_id
stringlengths
4
20
001-5
ambig
true
retails
Report the total revenue for each nation in 1995.
[{"id":"A","phrase":"total revenue","type":"finite","ambiguity_type":"semantic_computation","interpr(...TRUNCATED)
[{"id":"GQRY-A.0-B.0-C.0-D.0","query":"WITH revenue AS (\nSELECT n.n_nationkey,\nn.n_name,\nCOALESCE(...TRUNCATED)
GQRY-A.1-B.1-C.0-D.3
026-0
ambig
true
professional_basketball
For each team that Marcus Williams played for, compute his aggregate field goal percentage.
[{"id":"A","phrase":"Marcus Williams","type":"finite","ambiguity_type":"semantic_value","interpretat(...TRUNCATED)
[{"id":"GQRY-A.0-B.0-C.0","query":"SELECT\nt.tmID,\nt.name,\nCASE\nWHEN SUM(pt.fgAttempted) = 0 THEN(...TRUNCATED)
GQRY-A.0-B.1-C.0
036-0
ambig
true
github_repos
"For each repository with a GPL license, count the number of merged PRs created in the period of 202(...TRUNCATED)
[{"id":"A","phrase":"GPL license","type":"finite","ambiguity_type":"semantic_value","interpretations(...TRUNCATED)
[{"id":"GQRY-A.0-B.0","query":"WITH gpl_repos AS (\nSELECT grl.repo_name, grl.license\nFROM GITHUB_R(...TRUNCATED)
GQRY-A.1-B.0
058-2
ambig
true
financial
Compute the amount of deposits from December 1997 to end of 1998.
[{"id":"A","phrase":"amount","type":"finite","ambiguity_type":"semantic_computation","interpretation(...TRUNCATED)
[{"id":"GQRY-A.0-B.0","query":"SELECT COUNT(*)\nFROM trans\nWHERE operation = 'VKLAD'\nAND (date >= (...TRUNCATED)
GQRY-A.0-B.1
072-0
ambig
true
codebase_community
Find posts with many related posts.
[{"id":"A","phrase":"many","type":"infinite","ambiguity_type":"semantic_value","parameter_name":"cou(...TRUNCATED)
[{"id":"GQRY-B.0","query":"SELECT p.Id, p.Title, COUNT(DISTINCT pl.RelatedPostId) AS RelatedCount\nF(...TRUNCATED)
GQRY-B.2
087-0
ambig
true
student_club
Which events had a low budget remaining or high spending on food or advertisement?
[{"id":"A","phrase":"low budget remaining","type":"infinite","ambiguity_type":"semantic_value","para(...TRUNCATED)
[{"id":"GQRY-B.0","query":"WITH event_budget AS (\nSELECT\nlink_to_event AS event_id,\nSUM(remaining(...TRUNCATED)
GQRY-B.0
037-1
ambig
true
github_repos
"List all repository names with many issues opened in 2022, provide the corresponding issue count an(...TRUNCATED)
[{"id":"A","phrase":"many","type":"infinite","ambiguity_type":"semantic_value","parameter_name":"iss(...TRUNCATED)
[{"id":"GQRY-B.0","query":"WITH issue_counts AS (\nSELECT\njson_extract(repo, '$.name') AS repo_name(...TRUNCATED)
GQRY-B.0
002-4
ambig
true
retails
Count the number of suppliers that offer both air and rail shipping in America.
[{"id":"A","phrase":"suppliers that offer both air and rail shipping in America","type":"finite","am(...TRUNCATED)
[{"id":"GQRY-A.0-B.0-C.0","query":"WITH usa_suppliers_with_air AS (\nSELECT DISTINCT l.l_suppkey\nFR(...TRUNCATED)
GQRY-A.0-B.0-C.0
048-1
ambig
true
github_repos
"Count the number of repository IDs in YEAR_2023 that have public events and also received a large n(...TRUNCATED)
[{"id":"A","phrase":"public events","type":"finite","ambiguity_type":"semantic_computation","interpr(...TRUNCATED)
[{"id":"GQRY-A.0-B.0","query":null,"parameter_names":["pr_count_threshold"],"parameter_values":{"pr_(...TRUNCATED)
GQRY-A.1-B.0
005-0
ambig
true
retails
Which part has the highest price?
[{"id":"A","phrase":"highest price","type":"finite","ambiguity_type":"semantic_computation","interpr(...TRUNCATED)
[{"id":"GQRY-A.0-B.0","query":null,"parameter_names":[],"parameter_values":{},"exec_result":{"df":{"(...TRUNCATED)
GQRY-A.1-B.1
End of preview. Expand in Data Studio

ARCS

Ambiguity Resolution Corpus for SQL. 311 task instances across 6 databases.

Files

  • tasks.jsonl — one row per intended interpretation.
  • databases/sqlite/*.sqlite — databases referenced by the db field.
  • databases/column_meanings.json — column-level descriptions.

Loading

from datasets import load_dataset
ds = load_dataset("kiaw73b5o7a/arcs-bench", split="train")

Most rows have the gold SQL string (gold_queries[*].query) masked to null to avoid label leakage; a small sample retains it. Other gold_queries fields (e.g. exec_result) are kept on every row.

Downloads last month
456