As text-to-image models become increasingly prevalent, ensuring their equitable performance across diverse cultural contexts is critical. Efforts to mitigate cross-cultural biases have been hampered by trade-offs, including a loss in performance, factual inaccuracies, or offensive outputs.Despite widespread recognition of these challenges, an inability to reliably measure these biases has stalled progress. To address this gap, we introduce CAIRE (https://github.com/siddharthyayavaram/CAIRE), an evaluation metric that assesses the degree of cultural relevance of an image, given a user-defined set of labels. Our framework grounds entities and concepts in the image to a knowledge base and uses factual information to give independent graded judgments for each culture label.On a manually curated dataset of culturally salient but rare items built using language models, CAIRE surpasses all baselines by 22% F1 points. Additionally, we construct two datasets for culturally universal concepts, one comprising of T2I generated outputs and another retrieved from naturally-occurring data. CAIRE achieves Pearson’s correlations of 0.56 and 0.66 with human ratings on these sets, based on a 5-point Likert scale of cultural relevance. This demonstrates its strong alignment with human judgment across diverse image sources.

An idiom refers to a special type of multi-word expression whose meaning is figurative and cannot be deduced from the literal interpretation of its components. Idioms are prevalent in almost all languages and text genres, necessitating explicit handling by comprehensive NLP systems. Such phrases are referred to as Potentially Idiomatic Expressions (PIEs) and automatically identifying them in text is a challenging task. In this paper, we propose using a BERT-based model fine-tuned with custom objectives, to improve the accuracy of detecting PIEs in text. Our custom loss functions capture two important properties (word cohesion and language translation) to distinguish PIEs from non-PIEs. We conducted several experiments on 7 datasets and showed that incorporating custom objectives while training the model leads to substantial gains. Our models trained using this approach also have better sequence accuracy over DISC, a state-of-the-art PIE detection technique, along with good transfer capabilities.