Yusuke Ide, Joshua Tanner, Adam Nohejl, Jacob Hoffman, Justin Vasselli, Hidetaka Kamigaito, Taro Watanabe. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Large-scale Vision-Language Models (LVLMs) process both images and text, excelling in multimodal tasks such as image captioning and description generation. However, while these models excel at generating factual content, their ability to generate and evaluate texts reflecting perspectives on the same image, depending on the context, has not been sufficiently explored. To address this, we propose IRR: Image Review Rank, a novel evaluation framework designed to assess critic review texts from multiple perspectives. IRR evaluates LVLMs by measuring how closely their judgments align with human interpretations. We validate it using a dataset of images from 15 categories, each with five critic review texts and annotated rankings in both English and Japanese, totaling over 2,000 data instances. Our results indicate that, although LVLMs exhibited consistent performance across languages, their correlation with human annotations was insufficient, highlighting the need for further advancements. These findings highlight the limitations of current evaluation methods and the need for approaches that better capture human reasoning in Vision & Language tasks.

Yusuke Ide, Yuto Nishida, Justin Vasselli, Miyu Oba, Yusuke Sakai, Hidetaka Kamigaito, Taro Watanabe. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers). 2025.

In semiconductor manufacturing, it is required to detect anomalies which cause expensive defects. In recent years, Generative Adversarial Networks (GANs) have played a big role in anomaly detection. This study aims to detect anomalies by analyzing sensor data using a GAN when multivariate time series of sensor data are given. Our GAN could detect anomalies which cannot be detected visually. Experimental results indicated that an attention mechanism could tell us important sensors in detecting anomalies.

It is well known that many real world networks have a power-law degree distribution (the scale-free property). However, there are no rigorous results for continuous-time quantum walks on such realistic graphs. In this paper, we analyse the space–time behaviour of continuous-time quantum walks and random walks on the threshold network model, which is a reasonable candidate model having the scale-free property. We show that the quantum walker exhibits localisation at the starting point, although the random walker tends to spread uniformly.