Gagan Mundada, Rohan Surana, Nandhini Swaminathan, Bodhisattwa Prasad Majumder, Junda Wu, Julian McAuley, Zhouhang Xie. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Recent advances in Multimodal Large Language Models (MLLMs) have demonstrated impressive capabilities across various visionlanguage tasks.However, their reasoning abilities in the multimodal symbolic music domain remain largely unexplored.We introduce Wild-Score, the first in-the-wild multimodal symbolic music reasoning and analysis benchmark, designed to evaluate MLLMs' capacity to interpret real-world music scores and answer complex musicological queries.Each instance in WildScore is sourced from genuine musical compositions and accompanied by authentic user-generated questions and discussions, capturing the intricacies of practical music analysis.To facilitate a comprehensive evaluation, we propose a systematic taxonomy, comprising both high-level and fine-grained musicological ontologies.Furthermore, we frame complex music reasoning as multiple-choice question answering, enabling controlled and scalable assessment of MLLMs' symbolic music understanding.Empirical benchmarking of state-ofthe-art MLLMs on WildScore reveals intriguing patterns in their visual-symbolic reasoning, uncovering both promising directions and persistent challenges for MLLMs in symbolic music reasoning and analysis.We release the dataset 1 and code 2 .

Node localization in time-varying Internet of Things (IoT) networks is an essential problem due to increased delay and poor Signal-to-Interference plus Noise Ratio (SINR) at the Base Station (BS). To improve the received signal strength at the BS, Reconfigurable Intelligent Surface (RIS) has recently been used between transmitter and receiver. Additionally, novel phase prediction methods and optimal weight assignment frameworks have been proposed over RIS and BSs, respectively. Nevertheless, these methods suffer from poor performance due to their heuristic approach, resulting in more time consumption and poor SINR. Motivated by the aforementioned challenges, we propose a novel node localization method over a RIS-assisted time-varying IoT network using Machine Learning (ML) frameworks in this work. Firstly, the method computes the optimal phase configuration over the RIS corresponding to each element using coeff2phaseNN, which has been trained on channel coefficients among the transmitter, receiver, and RIS. Subsequently, the weight of the individual antenna element at the BS is optimized using the proposed VectorSync model. The results confirm that the coeff2phaseNN method demonstrates a reduction of 89.79% in total MSE loss compared to the Artificial Neural Network-RIS (ANN-RIS) method. Additionally, it demonstrates a 71.04% reduction in the absolute RIS phase prediction deviation from the optimal phase compared to the ANN-RIS method. Moreover, the proposed VectorSync method attains a 79.28% and 92.29% reduction in time required for optimal weight assignment compared to the Bartlett and Capon methods, respectively. Finally, the Localization Error(LR) using the proposed method is compared to conventional methods in a time-varying experimental scenario and found to be the minimum, i.e., 6.156%.