CLASS-IT: Conversational and Lecture-Aligned Small-Scale Instruction Tuning for BabyLMs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Capone, Luca, Bondielli, Alessandro, Lenci, Alessandro
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908619001298944
author Capone, Luca
Bondielli, Alessandro
Lenci, Alessandro
author_facet Capone, Luca
Bondielli, Alessandro
Lenci, Alessandro
contents This work investigates whether small-scale LMs can benefit from instruction tuning. We compare conversational and question-answering instruction tuning datasets, applied either in a merged or sequential curriculum, using decoder-only models with 100M and 140M parameters. Evaluation spans both fine-tuning (SuperGLUE) and zero-shot (BLiMP, EWoK, WUGs, entity tracking, and psycholinguistic correlation) settings. Results show that instruction tuning yields small but consistent gains in fine-tuning scenarios, with sequential curricula outperforming merged data; however, improvements do not consistently transfer to zero-shot tasks, suggesting a trade-off between interaction-focused adaptation and broad linguistic generalization. These results highlight both the potential and the constraints of adapting human-inspired learning strategies to low-resource LMs, and point toward hybrid, curriculum-based approaches for enhancing generalization under ecological training limits.
format Preprint
id arxiv_https___arxiv_org_abs_2510_25364
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CLASS-IT: Conversational and Lecture-Aligned Small-Scale Instruction Tuning for BabyLMs
Capone, Luca
Bondielli, Alessandro
Lenci, Alessandro
Computation and Language
This work investigates whether small-scale LMs can benefit from instruction tuning. We compare conversational and question-answering instruction tuning datasets, applied either in a merged or sequential curriculum, using decoder-only models with 100M and 140M parameters. Evaluation spans both fine-tuning (SuperGLUE) and zero-shot (BLiMP, EWoK, WUGs, entity tracking, and psycholinguistic correlation) settings. Results show that instruction tuning yields small but consistent gains in fine-tuning scenarios, with sequential curricula outperforming merged data; however, improvements do not consistently transfer to zero-shot tasks, suggesting a trade-off between interaction-focused adaptation and broad linguistic generalization. These results highlight both the potential and the constraints of adapting human-inspired learning strategies to low-resource LMs, and point toward hybrid, curriculum-based approaches for enhancing generalization under ecological training limits.
title CLASS-IT: Conversational and Lecture-Aligned Small-Scale Instruction Tuning for BabyLMs
topic Computation and Language
url https://arxiv.org/abs/2510.25364