Towards Efficient CoT Distillation: Self-Guided Rationale Selector for Better Performance with Fewer Rationales

JianZhi Yan, Le Liu, Youcheng Pan, Shiwei Chen, Yang Xiang, Buzhou Tang


Abstract
CoT distillation is critical for enhancing small language models’ (SLMs) reasoning by transferring multi-step reasoning capability from the larger teacher models. However, existing work underestimates the importance of rationale quality, focusing primarily on data quantity, which may result in transferring noisy or incorrect information to the student model. To address the above issues, we proposed Model-Oriented Rationale Selection Distillation (MoRSD), which can discern and select high quality rationales for distillation. We further propose a Rationale Difficulty (RD) metric to measure the ability of the student model to generate the correct answer under a given rationale. Compared to the baseline, we achieved 4.6% average accuracy improvement on seven datasets over three tasks, using fewer rationales by controlling their accuracy, diversity, and difficulty. Our results reveal that a small portion of the high quality rationales can enhance the reasoning ability of student models than the entire dataset. Our method promises to be a possible solution for efficient CoT distillation. Our code will be released in https://github.com/Leon221220/MoRSD.
Anthology ID:
2025.findings-emnlp.413
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2025
Month:
November
Year:
2025
Address:
Suzhou, China
Editors:
Christos Christodoulopoulos, Tanmoy Chakraborty, Carolyn Rose, Violet Peng
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
7818–7835
Language:
URL:
https://aclanthology.org/2025.findings-emnlp.413/
DOI:
Bibkey:
Cite (ACL):
JianZhi Yan, Le Liu, Youcheng Pan, Shiwei Chen, Yang Xiang, and Buzhou Tang. 2025. Towards Efficient CoT Distillation: Self-Guided Rationale Selector for Better Performance with Fewer Rationales. In Findings of the Association for Computational Linguistics: EMNLP 2025, pages 7818–7835, Suzhou, China. Association for Computational Linguistics.
Cite (Informal):
Towards Efficient CoT Distillation: Self-Guided Rationale Selector for Better Performance with Fewer Rationales (Yan et al., Findings 2025)
Copy Citation:
PDF:
https://aclanthology.org/2025.findings-emnlp.413.pdf
Checklist:
 2025.findings-emnlp.413.checklist.pdf