
Transformer models, including GPT-4 and fine-tuned BERT variants, exhibit limited generalisability to out-of-distribution perturbations in mathematical reasoning tasks despite strong in-distribution performance.


coremsg

evaluating-the-generalisability-of-transformer-models-in-mathematical-reasoning-tasks


Evaluating the Generalisability of Transformer Models in Mathematical Reasoning Tasks


title_rewrite


Mathematical language processing requires sophisticated methods to extract information, reason over mathematical elements, and produce real-world problem solutions. Recent research has advanced key components in this direction, including transformer-based language models, graph-based representations, and multi-modal encoding approaches.


advancing-mathematical-language-processing-insights-and-trends-from-recent-research


Advancing Mathematical Language Processing: Insights and Trends from Recent Research
