From 9b7fb4bcee63f6276c4af2da962ecae02dc07a98 Mon Sep 17 00:00:00 2001
From: =?utf8?q?J=C3=A9r=C3=B4me=20Benoit?= <jerome.benoit@piment-noir.org>
Date: Sat, 11 Oct 2025 21:18:58 +0200
Subject: [PATCH] fix(reforcexy): use a stable reward metric for optuna best
 trial 		selection
MIME-Version: 1.0
Content-Type: text/plain; charset=utf8
Content-Transfer-Encoding: 8bit

Signed-off-by: Jérôme Benoit <jerome.benoit@piment-noir.org>
---
 ReforceXY/user_data/freqaimodels/ReforceXY.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/ReforceXY/user_data/freqaimodels/ReforceXY.py b/ReforceXY/user_data/freqaimodels/ReforceXY.py
index 503551e..1388032 100644
--- a/ReforceXY/user_data/freqaimodels/ReforceXY.py
+++ b/ReforceXY/user_data/freqaimodels/ReforceXY.py
@@ -1232,7 +1232,7 @@ class ReforceXY(BaseReinforcementLearningModel):
         if self.optuna_eval_callback.is_pruned:
             raise TrialPruned()
 
-        return self.optuna_eval_callback.last_mean_reward
+        return self.optuna_eval_callback.best_mean_reward
 
     def close_envs(self) -> None:
         """
-- 
2.43.0