paper-with-me

홈 › Papers

Shortcut Solutions Learned by Transformers Impair Continual Compositional Reasoning

2026-05-06 · William T. Redman, Erik C. Johnson, Brian Robinson arxiv

Identifying and exploiting common features across domains is at the heart of the human ability to make analogies, and is believed to be crucial for the ability to continually learn. To do this successfully, general and flexible computational strategies must be developed. While the extent to which Transformer neural network models can perform compositional reasoning has been the subject of intensive recent investigation, little work has been done to systematically understand how well these models can leverage their representations to learn new, related experiences. To address this gap, we expand the previously developed Learning Equality and Group Operations (LEGO) framework to a continual learning (CL) setting ("continual LEGO"). Using this continual LEGO experimental paradigm, we study the capability of feedforward and recurrent Transformer models to perform CL. We find that BERT, a canonical feedforward Transformer model, learns shortcut solutions that limits its ability to generalize and prevents strong forward transfer to new experiences. In contrast, we find evidence supporting the hypothesis that ALBERT, a recurrent version of BERT, learns a For loop-esque solution, which leads to better CL performance. When applying BERT and ALBERT models to a CL setting that requires composition across experiences, we find that both model families fail. Our investigation suggests that ALBERT models can have their performance drop rescued by use of training strategies that combine data across experiences, but this is not true for BERT models, where a detrimental shortcut solution becomes entrenched with initial training. Our results demonstrate that the recurrent ALBERT model may have an inductive bias better suited for CL and motivate future investigation of the interplay between Transformer architecture and computational solutions that emerge in modern models and tasks.

📄 PDF Abstract BibTeX arXiv:2605.05495

Code (0)

등록된 구현이 없습니다.

Tasks

Continual Learning

Similar Papers 제목 키워드 기반

Transformers Learn Shortcuts to Automata

2022-10-19 · Bingbin Liu, Jordan T. Ash, Surbhi Goel, Akshay Krishnamurthy 외

Algorithmic reasoning requires capabilities which are most naturally understood through recurrent models of computation, like the Turing machine. However, Transformer models, while lacking recurrence, are able to perform…

Which Shortcut Solution Do Question Answering Models Prefer to Learn?

2022-11-29 · Kazutoshi Shinoda, Saku Sugawara, Akiko Aizawa

Question answering (QA) models for reading comprehension tend to learn shortcut solutions rather than the solutions intended by QA datasets. QA models that have learned shortcut solutions can achieve human-level performa…

Multiple-choiceQuestion AnsweringReading Comprehension

Diagnosing Shortcut-Induced Rigidity in Continual Learning: The Einstellung Rigidity Index (ERI)

2025-10-01 · Kai Gu, Weishi Shi arxiv

Deep neural networks frequently exploit shortcut features, defined as incidental correlations between inputs and labels without causal meaning. Shortcut features undermine robustness and reduce reliability under distribu…

Continual Learning

Do ImageNet-trained models learn shortcuts? The impact of frequency shortcuts on generalization

2025-03-05 · CVPR 2025 1 · Shunxin Wang, Raymond Veldhuis, Nicola Strisciuglio

Frequency shortcuts refer to specific frequency patterns that models heavily rely on for correct classification. Previous studies have shown that models trained on small image datasets often exploit such shortcuts, poten…

Overcoming Shortcut Learning in a Target Domain by Generalizing Basic Visual Factors from a Source Domain

2022-07-20 · Piyapat Saranrittichai, Chaithanya Kumar Mummadi, Claudia Blaiotta, Mauricio Munoz 외

Shortcut learning occurs when a deep neural network overly relies on spurious correlations in the training dataset in order to solve downstream tasks. Prior works have shown how this impairs the compositional generalizat…