Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion pyproject.toml
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@ dependencies = [
"jinja2",
"huggingface-hub>=0.20",
"pyarrow>=14.0",
"kernelguard>=0.1.1",
"kernelguard==0.3.1",
]

[project.optional-dependencies]
Expand Down
28 changes: 28 additions & 0 deletions src/libkernelbot/leaderboard_db.py
Original file line number Diff line number Diff line change
Expand Up @@ -883,6 +883,13 @@ def get_leaderboard_submissions(
AND NOT r.secret
AND r.score IS NOT NULL
AND r.passed
AND s.status <> 'hacked'
AND NOT EXISTS (
SELECT 1
FROM leaderboard.submission_job_status sjs
WHERE sjs.submission_id = s.id
AND sjs.status = 'hacked'
)
AND s.user_id = %s
AND EXISTS (
SELECT 1
Expand Down Expand Up @@ -922,6 +929,13 @@ def get_leaderboard_submissions(
JOIN leaderboard.user_info ui ON s.user_id = ui.id
WHERE l.name = %s AND r.runner = %s AND NOT r.secret
AND r.score IS NOT NULL AND r.passed
AND s.status <> 'hacked'
AND NOT EXISTS (
SELECT 1
FROM leaderboard.submission_job_status sjs
WHERE sjs.submission_id = s.id
AND sjs.status = 'hacked'
)
AND EXISTS (
SELECT 1
FROM leaderboard.runs sr
Expand Down Expand Up @@ -1441,6 +1455,13 @@ def get_leaderboard_submission_count(
AND NOT r.secret
AND r.score IS NOT NULL
AND r.passed
AND s.status <> 'hacked'
AND NOT EXISTS (
SELECT 1
FROM leaderboard.submission_job_status sjs
WHERE sjs.submission_id = s.id
AND sjs.status = 'hacked'
)
AND s.user_id = %s
AND EXISTS (
SELECT 1
Expand Down Expand Up @@ -1472,6 +1493,13 @@ def get_leaderboard_submission_count(
AND NOT r.secret
AND r.score IS NOT NULL
AND r.passed
AND s.status <> 'hacked'
AND NOT EXISTS (
SELECT 1
FROM leaderboard.submission_job_status sjs
WHERE sjs.submission_id = s.id
AND sjs.status = 'hacked'
)
AND EXISTS (
SELECT 1
FROM leaderboard.runs sr
Expand Down
46 changes: 46 additions & 0 deletions tests/test_leaderboard_db.py
Original file line number Diff line number Diff line change
Expand Up @@ -385,6 +385,52 @@ def test_leaderboard_submission_ranked(database, submit_leaderboard):
]


def test_hacked_submissions_are_hidden_from_leaderboard_rankings(database, submit_leaderboard):
submit_time = datetime.datetime.now(tz=datetime.timezone.utc)

with database as db:
submission_status_hacked = db.create_submission(
"submit-leaderboard", "submission_status_hacked.py", 5, "fast", submit_time, user_name="user5"
)
_create_submission_run(
db, submission_status_hacked, mode="leaderboard", runner="A100", score=1.0
)
_create_submission_run(
db, submission_status_hacked, mode="leaderboard", secret=True, runner="A100"
)
db.mark_submission_hacked(submission_status_hacked, error="blocked")

job_status_hacked = db.create_submission(
"submit-leaderboard", "job_status_hacked.py", 6, "fast", submit_time, user_name="user6"
)
_create_submission_run(db, job_status_hacked, mode="leaderboard", runner="A100", score=0.5)
_create_submission_run(db, job_status_hacked, mode="leaderboard", secret=True, runner="A100")
db.mark_submission_done(job_status_hacked)
db.upsert_submission_job_status(job_status_hacked, "hacked", "blocked")

valid = db.create_submission(
"submit-leaderboard", "valid.py", 7, "valid", submit_time, user_name="user7"
)
_create_submission_run(db, valid, mode="leaderboard", runner="A100", score=2.0)
_create_submission_run(db, valid, mode="leaderboard", secret=True, runner="A100")
db.mark_submission_done(valid)

with database as db:
ranked = db.get_leaderboard_submissions("submit-leaderboard", "A100")
assert [row["submission_id"] for row in ranked] == [valid]

assert db.get_leaderboard_submissions("submit-leaderboard", "A100", "5") == []
assert db.get_leaderboard_submissions("submit-leaderboard", "A100", "6") == []
assert [row["submission_id"] for row in db.get_leaderboard_submissions(
"submit-leaderboard", "A100", "7"
)] == [valid]

assert db.get_leaderboard_submission_count("submit-leaderboard", "A100") == 1
assert db.get_leaderboard_submission_count("submit-leaderboard", "A100", "5") == 0
assert db.get_leaderboard_submission_count("submit-leaderboard", "A100", "6") == 0
assert db.get_leaderboard_submission_count("submit-leaderboard", "A100", "7") == 1


def test_failed_secret_run_hides_submission_from_rankings(database, submit_leaderboard):
submit_time = datetime.datetime.now(tz=datetime.timezone.utc)
failed_secret = dataclasses.replace(sample_run_result(), passed=False)
Expand Down
8 changes: 4 additions & 4 deletions uv.lock

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

Loading