Metraipolis › archive › dc_07fa0f05afb9
Fractal basins trap latent reasoning — arXiv:2609.04963
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.04963
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T22:58:26.142Z
- SHA-256 of the archived text
962cdef18251f09f454cb9b302b31632a4945f6f6a66c4b47f71487dfd99ab87- Stable URL of this page
https://metraipolis.com/documento/dc_07fa0f05afb9
How to cite
Fractal basins trap latent reasoning — arXiv:2609.04963 — arXiv (en), https://arxiv.org/abs/2609.04963 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_07fa0f05afb9 (Metraipolis archive dc_07fa0f05afb9, sha256 962cdef18251f09f454cb9b302b31632a4945f6f6a66c4b47f71487dfd99ab87)
BibTeX
@misc{dc07fa0f05afb9,
title = {Fractal basins trap latent reasoning — arXiv:2609.04963},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.04963},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_07fa0f05afb9; sha256 962cdef18251f09f454cb9b302b31632a4945f6f6a66c4b47f71487dfd99ab87},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.04963. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Reasoning allows artificial intelligence models to revisit and correct their mistakes, enabling recent frontier advances in mathematical theorem solving, software engineering, and autonomous task planning. Reasoning models are widely observed to reason for longer on harder tasks, but the general mechanism responsible for these slowdowns is unknown. Here, we show that reasoning models exhibit transient chaos, a physical consequence of the computational complexity of difficult tasks. As a consequence, we show that diverse leading reasoning models are dynamical systems with fractal basins, with fractality increasing with task difficulty across diverse tasks like Sudoku and maze solving, visual puzzles, and mathematical logic. We show that transient chaos emerges due to reasoning becoming trapped for extended durations near saddle points, which we show correspond to nearly-correct attempted solutions of the underlying problem. Our results show that reasoning slowdowns are an inevitable consequence of problem hardness in modern artificial intelligence models, and establish reasoning traces as a rich new class of dynamical system. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-07 in cs.LG). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.04963 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-16T22:31:32.200Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-07?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).