We describe our submission to SemEval-2026 Task 2 (Subtask 1), which asks systems to predict continuous Valence and Arousal scores from ecological diary texts. We fine-tune RoBERTa-base with a single linear regression head, treating each essay independently. Our system scores rcomposite of .679 (Valence) and .466 (Arousal) on the official test set, placing 4th on the Subtask 1 leaderboard.