From 53497642985aa1184cfa63378612b6db10599782 Mon Sep 17 00:00:00 2001 From: Douglas Yang Date: Tue, 6 Jan 2026 13:18:43 -0800 Subject: [PATCH] fix: kimi k2 thinking accuracy threshold change (#16585) --- test/registered/8-gpu-models/test_kimi_k2.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/test/registered/8-gpu-models/test_kimi_k2.py b/test/registered/8-gpu-models/test_kimi_k2.py index dd873d695..079dbcaf3 100644 --- a/test/registered/8-gpu-models/test_kimi_k2.py +++ b/test/registered/8-gpu-models/test_kimi_k2.py @@ -42,7 +42,7 @@ class TestKimiK2Unified(unittest.TestCase): run_combined_tests( models=variants, test_name="Kimi-K2-Thinking Unified", - accuracy_params=AccuracyTestParams(dataset="gsm8k", baseline_accuracy=0.95), + accuracy_params=AccuracyTestParams(dataset="gsm8k", baseline_accuracy=0.94), performance_params=PerformanceTestParams( profile_dir="performance_profiles_kimi_k2_thinking", ),