Zhongyuan Peng, Yifan Yao, Kaijing Ma, Shuyue Guo, Yizhe Li, Yichi Zhang, Chenchen Zhang, Yifan Zhang, Zhouliang Yu, Luming Li, Minghao Liu, Yihang Xia, Jiawei Shen, Yuchen Wu, Yixin Cao, Zhaoxiang Zhang, Wenhao Huang, Jiaheng Liu, Ge Zhang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

In this paper, we explore the application of Large Language Models (LLMs) to the pre-training of music. While the prevalent use of MIDI in music modeling is well-established, our findings suggest that LLMs are inherently more compatible with ABC Notation, which aligns more closely with their design and strengths, thereby enhancing the model's performance in musical composition. To address the challenges associated with misaligned measures from different tracks during generation, we propose the development of a $\underline{S}$ynchronized $\underline{M}$ulti-$\underline{T}$rack ABC Notation ($\textbf{SMT-ABC Notation}$), which aims to preserve coherence across multiple musical tracks. Our contributions include a series of models capable of handling up to 8192 tokens, covering 90\% of the symbolic music data in our training set. Furthermore, we explore the implications of the $\underline{S}$ymbolic $\underline{M}$usic $\underline{S}$caling Law ($\textbf{SMS Law}$) on model performance. The results indicate a promising research direction in music generation, offering extensive resources for further research through our open-source contributions.

Reconfigurable intelligent surface (RIS) is highly envisaged as a promising technology in millimeter wave (mmWave) communication systems for its capability of restructuring the wireless communication environment and mitigating the severe signal blockage. However, passive RIS beamforming design is still a challenge due to the non-convex property of the problem. This work presents a deep learning (DL) based codebook classification beam search algorithm, comprising a convolutional neural network (CNN) based codebook searcher and a mapper. The searcher determines the optimal codeword position range index based on the channel state information (CSI), while the mapper selects the optimal codeword based on this index. The proposed network is trained both with the DeepMIMO dataset and the Saleh-Valenzuela channel model, respectively, and imperfect CSI is utilized to enhance the system robustness. Simulation results demonstrate that the proposed algorithm significantly improves the beam search efficiency.