A Psychology-based Unified Dynamic Framework for Curriculum Learning

Guangyu Meng, Qingkai Zeng, John P. Lalor and Hong Yu published a Psychology-based Unified Dynamic Framework for Curriculum Learning (PUDF) in Computational Linguistics Volume 52, Issue 3 (September 2026, pages 1001–1049, MIT Press). PUDF quantifies training-data difficulty by applying Item Response Theory to responses from Artificial Crowds, yielding global, model-independent and interpretable difficulty values, and uses a Dynamic Data Selection via Model Ability Estimation (DDS-MAE) strategy to schedule how much data is fed at each training step. The authors report that fine-tuning pre-trained large language models with PUDF produced higher accuracy and faster convergence across a suite of benchmark datasets than standard fine-tuning and state-of-the-art curriculum learning methods.

Abstract Directly learning from examples of varying difficulty levels is often challenging for both humans and machine learning models. A more effective strategy involves exposing learners to examples in a progressive order from easy to difficult. Curriculum Learning CL has been proposed to implement this strategy in machine learning model training. However, two key challenges persist in CL framework design: defining the difficulty of training data and determining the appropriate amount of data to input at each training step. Drawing inspiration from psychometrics, this article presents a Psychology-based Unified Dynamic Framework for Curriculum Learning PUDF . We quantify the difficulty of training data by applying Item Response Theory IRT to responses from Artificial Crowds AC . This theory-driven IRT-AC approach leads to global i.e., model-independent and interpretable difficulty values. Leveraging IRT, we propose a training strategy, Dynamic Data Selection via Model Ability Estimation DDS-MAE , to schedule the appropriate amount of data during model training. Because our difficulty labeling and model ability estimation are based on a consistent theory, namely, IRT, their values are comparable within the same scope, potentially leading to aligned training data selection and faster convergence compared with the other CL methods. Experimental results demonstrate that fine-tuning pre-trained large language models with PUDF leads to higher accuracy and faster convergence on a suite of benchmark datasets compared with standard fine-tuning and state-of-the-art CL methods. Ablation studies and downstream analyses further validate the impact of PUDF for CL. - Anthology ID: - 2026.cl-3.4 - Volume: - Computational Linguistics, Volume 52, Issue 3 - September 2026 https://aclanthology.org/volumes/2026.cl-3/ - Month: - September - Year: - 2026 - Address: - Cambridge, MA - Venue: - CL https://aclanthology.org/venues/cl/ - SIG: - Publisher: - MIT Press - Note: - Pages: - 1001–1049 - Language: - URL: - https://aclanthology.org/2026.cl-3.4/ https://aclanthology.org/2026.cl-3.4/ - DOI: - 10.1162/coli.a.584 https://doi.org/10.1162/coli.a.584 - Cite ACL : - Guangyu Meng, Qingkai Zeng, John P. Lalor, and Hong Yu. 2026. A Psychology-based Unified Dynamic Framework for Curriculum Learning https://aclanthology.org/2026.cl-3.4/ . Computational Linguistics , 52 3 :1001–1049. - Cite Informal : - A Psychology-based Unified Dynamic Framework for Curriculum Learning https://aclanthology.org/2026.cl-3.4/ Meng et al., CL 2026 - PDF: - https://aclanthology.org/2026.cl-3.4.pdf https://aclanthology.org/2026.cl-3.4.pdf