MENTOR: A Reinforcement Learning Framework for Enabling Tool Use in Small Models via Teacher-Optimized Rewards

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Choi, ChangSu, Song, Hoyun, Kim, Dongyeon, Jung, WooHyeon, Cho, Minkyung, Park, Sunjin, Bae, NohHyeob, Yu, Seona, Lim, KyungTae
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912673490272256
author Choi, ChangSu
Song, Hoyun
Kim, Dongyeon
Jung, WooHyeon
Cho, Minkyung
Park, Sunjin
Bae, NohHyeob
Yu, Seona
Lim, KyungTae
author_facet Choi, ChangSu
Song, Hoyun
Kim, Dongyeon
Jung, WooHyeon
Cho, Minkyung
Park, Sunjin
Bae, NohHyeob
Yu, Seona
Lim, KyungTae
contents Distilling the tool-using capabilities of large language models (LLMs) into smaller, more efficient small language models (SLMs) is a key challenge for their practical application. The predominant approach, supervised fine-tuning (SFT), suffers from poor generalization as it trains models to imitate a static set of teacher trajectories rather than learn a robust methodology. While reinforcement learning (RL) offers an alternative, the standard RL using sparse rewards fails to effectively guide SLMs, causing them to struggle with inefficient exploration and adopt suboptimal strategies. To address these distinct challenges, we propose MENTOR, a framework that synergistically combines RL with teacher-guided distillation. Instead of simple imitation, MENTOR employs an RL-based process to learn a more generalizable policy through exploration. In addition, to solve the problem of reward sparsity, it uses a teacher's reference trajectory to construct a dense, composite teacher-guided reward that provides fine-grained guidance. Extensive experiments demonstrate that MENTOR significantly improves the cross-domain generalization and strategic competence of SLMs compared to both SFT and standard sparse-reward RL baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2510_18383
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle MENTOR: A Reinforcement Learning Framework for Enabling Tool Use in Small Models via Teacher-Optimized Rewards
Choi, ChangSu
Song, Hoyun
Kim, Dongyeon
Jung, WooHyeon
Cho, Minkyung
Park, Sunjin
Bae, NohHyeob
Yu, Seona
Lim, KyungTae
Computation and Language
Artificial Intelligence
Distilling the tool-using capabilities of large language models (LLMs) into smaller, more efficient small language models (SLMs) is a key challenge for their practical application. The predominant approach, supervised fine-tuning (SFT), suffers from poor generalization as it trains models to imitate a static set of teacher trajectories rather than learn a robust methodology. While reinforcement learning (RL) offers an alternative, the standard RL using sparse rewards fails to effectively guide SLMs, causing them to struggle with inefficient exploration and adopt suboptimal strategies. To address these distinct challenges, we propose MENTOR, a framework that synergistically combines RL with teacher-guided distillation. Instead of simple imitation, MENTOR employs an RL-based process to learn a more generalizable policy through exploration. In addition, to solve the problem of reward sparsity, it uses a teacher's reference trajectory to construct a dense, composite teacher-guided reward that provides fine-grained guidance. Extensive experiments demonstrate that MENTOR significantly improves the cross-domain generalization and strategic competence of SLMs compared to both SFT and standard sparse-reward RL baselines.
title MENTOR: A Reinforcement Learning Framework for Enabling Tool Use in Small Models via Teacher-Optimized Rewards
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2510.18383