change to 1 run

xzrderek · xzrderek · commit 4aca2728e66b · 2026-02-16T16:29:53.000-08:00
diff --git a/eval_protocol/benchmarks/test_aime25.py b/eval_protocol/benchmarks/test_aime25.py
@@ -90,7 +90,7 @@ def aime2025_dataset_adapter(rows: List[Dict[str, Any]]) -> List[EvaluationRow]:
     rollout_processor=SingleTurnRolloutProcessor(),
     aggregation_method="mean",
     passed_threshold=0.8,
-    num_runs=2,
+    num_runs=1,
     max_concurrent_rollouts=16,
     mode="pointwise",
 )

Original file line number	Diff line number	Diff line change
`@@ -90,7 +90,7 @@ def aime2025_dataset_adapter(rows: List[Dict[str, Any]]) -> List[EvaluationRow]:`
`90`	`90`	`rollout_processor=SingleTurnRolloutProcessor(),`
`91`	`91`	`aggregation_method="mean",`
`92`	`92`	`passed_threshold=0.8,`
`93`		`- num_runs=2,`
	`93`	`+ num_runs=1,`
`94`	`94`	`max_concurrent_rollouts=16,`
`95`	`95`	`mode="pointwise",`
`96`	`96`	`)`