GPT-4 may score in the top 10% of the Bar Exam, but it still fails at true OOD reasoning. This 2025 psychological audit explains why AGI remains a challenge. submitted by /u/jorgebscomm [link] [comments]