From ea81e7c49d21e13074a3bc1a41bc052b615d4642 Mon Sep 17 00:00:00 2001 From: Shay Date: Wed, 17 Jun 2026 13:13:20 -0700 Subject: [PATCH] chore(gsm8k): polish monotonic baseline test hygiene --- tests/gsm8k_train_sample_baseline.py | 2 +- tests/test_gsm8k_frontier_report.py | 3 --- 2 files changed, 1 insertion(+), 4 deletions(-) diff --git a/tests/gsm8k_train_sample_baseline.py b/tests/gsm8k_train_sample_baseline.py index f0e54c28..aca3113b 100644 --- a/tests/gsm8k_train_sample_baseline.py +++ b/tests/gsm8k_train_sample_baseline.py @@ -28,4 +28,4 @@ def assert_monotonic_serving_counts(counts: dict[str, int]) -> None: total = counts["correct"] + counts["wrong"] + counts["refused"] assert total == TRAIN_SAMPLE_COUNT, ( f"expected {TRAIN_SAMPLE_COUNT} scored cases, got {total}" - ) \ No newline at end of file + ) diff --git a/tests/test_gsm8k_frontier_report.py b/tests/test_gsm8k_frontier_report.py index 1fde24bf..24789dc9 100644 --- a/tests/test_gsm8k_frontier_report.py +++ b/tests/test_gsm8k_frontier_report.py @@ -60,9 +60,6 @@ def test_committed_frontier_fixture_is_6_44_0_until_rebaselined(): def test_live_serving_meets_monotonic_capability_contract(): """Live train_sample: wrong=0 hard; counts monotonic; refusals carry reasons.""" - import re - from collections import Counter - from evals.gsm8k_math.train_sample.v1.runner import build_report cases_path = _REPO_ROOT / "evals/gsm8k_math/train_sample/v1/cases.jsonl"