LARK: Learnability-Grounded Trajectory Selection for Efficient Reasoning Distillation

Tianrun Yu, Kaixiang Zhao, Chih-Chun Chen, Amanda Hughes, Taylor W. Killian, Fenglong Ma, Weitong Zhang, Porter Jenkins, [0.5em] Brigham Young University The Pennsylvania State University University of North Carolina at Chapel Hill [0.4em] *Corresponding author: [email protected] · 2026

Read LARK: Learnability-Grounded Trajectory Selection for Efficient Reasoning Distillation by Tianrun Yu, Kaixiang Zhao, Chih-Chun Chen, Amanda Hughes, Taylor W. Killian, Fenglong Ma, Weitong Zhang, Porter Jenkins, [0.5em] Brigham Young University The Pennsylvania State University University of North Carolina at Chapel Hill [0.4em] *Corresponding author: [email protected] (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-30651.txt.

Introduction

Yu and colleagues study path selection for reasoning distillation, where teacher paths supervise a student model. Old methods rely on quality or trust, missing whether the student can learn a trajectory. Their LARK method selects trajectories the student learns efficiently while keeping generalization. At its core is a learnability factor for how fast the student's loss falls, estimated with a proxy and a policy balancing learnability and coverage. LARK beats data selection baselines across models and tasks, and its score predicts utility with faster loss drop.

Catalog record

CREATORTianrun Yu, Kaixiang Zhao, Chih-Chun Chen, Amanda Hughes, Taylor W. Killian, Fenglong Ma, Weitong Zhang, Porter Jenkins, [0.5em] Brigham Young University The Pennsylvania State University University of North Carolina at Chapel Hill [0.4em] *Corresponding author: [email protected]
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE163485
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.30651

Browse the full library · About the Tower of Babel