diff --git a/mteb/evaluation/evaluators/utils.py b/mteb/evaluation/evaluators/utils.py
index 8c8850a3c6..d6ffe535a3 100644
--- a/mteb/evaluation/evaluators/utils.py
+++ b/mteb/evaluation/evaluators/utils.py
@@ -398,7 +398,7 @@ def abstention_curve(
         Returns:
             abst_curve: Abstention curve of length `len(abstention_rates)`
         """
-        conf_scores_argsort = np.argsort(conf_scores)
+        conf_scores_argsort = np.argsort(conf_scores, stable=True)
         abst_curve = np.zeros(len(abstention_rates))
 
         for i, rate in enumerate(abstention_rates):
diff --git a/pyproject.toml b/pyproject.toml
index 65638d0281..aff9ff337e 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -26,7 +26,7 @@ classifiers = [
 requires-python = ">=3.9"
 dependencies = [
     "datasets>=2.19.0",
-    "numpy>=1.0.0,<2.0.0", # note: https://github.com/huggingface/datasets/issues/6980
+    "numpy>=1.0.0,<3.0.0",
     "requests>=2.26.0",
     "scikit_learn>=1.0.2",
     "scipy>=0.0.0",