Project Aletheia: Verifier-Guided Distillation of Backtracking for Small Language Models

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Dixit, Aradhya, Liang, Tianxi, Telang, Jai
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912836448419840
author Dixit, Aradhya
Liang, Tianxi
Telang, Jai
author_facet Dixit, Aradhya
Liang, Tianxi
Telang, Jai
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2601_14290
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Project Aletheia: Verifier-Guided Distillation of Backtracking for Small Language Models
Dixit, Aradhya
Liang, Tianxi
Telang, Jai
Computation and Language
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.
title Project Aletheia: Verifier-Guided Distillation of Backtracking for Small Language Models
topic Computation and Language
url https://arxiv.org/abs/2601.14290