IAAIThink Shallow, Solve Deep
Think Shallow, Solve Deep: Controlling Recurrent Dynamics for Reliable Test-Time Depth
Shows that a trained recurrent operator's finite-time dynamical regime, estimated as settling, marginal or drifting, predicts whether extra test-time iterations help, preserve or hurt answers, and gives a sufficient condition for depth safety.
Inside the method
[ Operator ] × r → regime: settling / marginal / drifting
Simplified conceptual schematic. Consult the paper for the complete architecture.
- Recurrence family
- Core recurrence
- Depth control
- Test-time depth beyond training
- KV / state strategy
- See paper
Reading note
Settling operators do not degrade with depth, and on Sudoku accuracy rises from 0.19 to 0.34 past the training horizon. A terminal fixed-point objective controls the regime; Huginn-3.5B falls in the non-settling family.
Sources checked 2026-10-28. This catalog does not imply independent reproduction.
Related works
Cite this work
@misc{thinkshallowsolvedeep2026,
title = {Think Shallow, Solve Deep: Controlling Recurrent Dynamics for Reliable Test-Time Depth},
author = {Ivan Viakhirev and Kirill Borodin and Amirah Almutairi and Serguei Barannikov and Maxim Abramov and Grach Mkrtchian},
year = {2026},
eprint = {2608.18222},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.18222}
}