Small Language Models (SLMs, under 10B parameters) are attractive for private, on-device deployment, yet they frequently fail on strict constraint-satisfaction problems due to linear, overconfident reasoning traces that do not recover from early mistakes. We introduce Verifier-Guided Distillation, a training protocol that transfers the process of error repair - explicit conflict detection and backtracking - rather than only correct final answers. By training a 7B model on verified reasoning traces that include mistakes and self-corrections, we show that latent verification behavior can emerge in small models, enabling them to occasionally stop, detect contradictions, and revise earlier assumptions.
@article{arxiv.2601.14290,
title = {Project Aletheia: Verifier-Guided Distillation of Backtracking for Small Language Models},
author = {Aradhya Dixit and Tianxi Liang and Jai Telang},
journal= {arXiv preprint arXiv:2601.14290},
year = {2026}
}