Researchers continue to question whether AI systems demonstrate genuine logical reasoning or merely exploit surface-level patterns to solve problems. While large reasoning models achieve impressive results on benchmarks, experts note that their generated thought processes are often disconnected from their internal logic.