Human vs random-AI (Eval 17)
Human vs random-AI (3-lens) . PsyProxy quantitative evaluation. The excerpts reflect a diverse range of personal experiences and challenges, primarily revolving around relationships , mental health , and personal growth . Themes of communication and emotional support recur throughout, emphasizing the importance of understanding and empathy in navigating difficult situations, whether in romantic partnerships, friendships, or family dynamics. The tone varies from supportive and informative to candid and reflective, with many texts offering practical advice for improving self-confidence , managing stress, and fostering healthy boundaries. Additionally, there are discussions on societal issues such as financial struggles and the impact of substance abuse , showcasing the complexity of human experiences and the nuanced conversations surrounding them. [Summary on 50 random texts by ChatGPT 4o Mini].
Reported holdout systems from the verified card
Binary classification uses FVE as the task-primary metric. Secondary columns keep the companion metrics visible so binary, ordinal, regression, and multiclass cards are not compared through one flattened score.