Comment on 'The Illusion of Thinking'
Three days later a rebuttal says the 'collapse' is mostly token-limit truncation, flawed grading and unsolvable puzzle instances.
Argues Hanoi runs exceed output limits (models say so), the grader conflates truncation with reasoning failure, and River Crossing is impossible for N>5. Asking for a generating function instead of a move list restores high accuracy. Became the standard counter-citation.
- Date
- Tuesday, 10 June 2025
- Lab
- Independent (A. Lawsen)
- Kind
- paper
- Access
- paper only
Single-author comment by A. Lawsen (arXiv v1 2025-06-10). A methodological counter-argument, not a replication of the original puzzles.
Sources
This record was checked against its sources on 6 October 2026. How we check