diff --git a/src/envs/textarena_env/rewards.py b/src/envs/textarena_env/rewards.py index bbd78b54..40d82a86 100644 --- a/src/envs/textarena_env/rewards.py +++ b/src/envs/textarena_env/rewards.py @@ -65,7 +65,7 @@ def extract_feedback_counts(feedback: str) -> Tuple[int, int]: if len(lines) < 2: return (0, 0) - for line in lines: + for line in reversed(lines): normalized = line.replace(" ", "") if normalized and all(c in "GYX" for c in normalized): green = normalized.count("G")