Search

We introduce Seed-Music, a suite of music generation systems capable of producing high-quality music with fine-grained style control. Our unified framework leverages both auto-regressive language modeling and diffusion approaches to support two key music creation workflows: controlled music generation and post-production editing. For controlled music generation, our system enables vocal music generation with performance controls from multi-modal inputs, including style descriptions, audio references, musical scores, and voice prompts. For post-production editing, it offers interactive tools for editing lyrics and vocal melodies directly in the generated audio. We encourage readers to listen to demo audio examples at https://team.doubao.com/seed-music "https://team.doubao.com/seed-music"., Comment: Seed-Music technical report, 20 pages, 5 figures

Decoding linguistic information from non-invasive brain signals using EEG has gained increasing research attention due to its vast applicational potential. Recently, a number of works have adopted a generative-based framework to decode electroencephalogram (EEG) signals into sentences by utilizing the power generative capacity of pretrained large language models (LLMs). However, this approach has several drawbacks that hinder the further development of linguistic applications for brain-computer interfaces (BCIs). Specifically, the ability of the EEG encoder to learn semantic information from EEG data remains questionable, and the LLM decoder's tendency to generate sentences based on its training memory can be hard to avoid. These issues necessitate a novel approach for converting EEG signals into sentences. In this paper, we propose a novel two-step pipeline that addresses these limitations and enhances the validity of linguistic EEG decoding research. We first confirm that word-level semantic information can be learned from EEG data recorded during natural reading by training a Conformer encoder via a masked contrastive objective for word-level classification. To achieve sentence decoding results, we employ a training-free retrieval method to retrieve sentences based on the predictions from the EEG encoder. Extensive experiments and ablation studies were conducted in this paper for a comprehensive evaluation of the proposed approach. Visualization of the top prediction candidates reveals that our model effectively groups EEG segments into semantic categories with similar meanings, thereby validating its ability to learn patterns from unspoken EEG recordings. Despite the exploratory nature of this work, these results suggest that our method holds promise for providing more reliable solutions for converting EEG signals into text.

Due to circuit failures, defective elements that cannot adaptively adjust the phase shifts of their impinging signals in a desired manner may exist on an intelligent reflecting surface (IRS). Traditional way to find these defective IRS elements requires a thorough diagnosis of all the circuits belonging to a huge number of IRS elements, which is practically challenging. In this paper, we will devise a novel approach under which a transmitter sends known pilot signals and a receiver localizes all the defective IRS elements just based on its over-the-air measurements reflected from the IRS. The key lies in the fact that the over-the-air measurements at the receiver side are functions of the set of defective IRS elements. Based on this observation, we propose a bisection based method to localize all the defective IRS elements. Specifically, at each time slot, we properly control the desired phase shifts of all the IRS elements such that half of the considered regime that is not useful to localize the defective elements can be found based on the received signals and removed. Via numerical results, it is shown that our proposed bisection method can exploit the over-the-air measurements to localize all the defective IRS elements quickly and accurately., Comment: accepted by 2024 IEEE Globecom

By Perelman's $\mathcal L$-geodesic theory, we study the blow-down solutions on a noncompact $\kappa$-noncollapsed steady gradient Ricci soliton $(M^n, g)$ $(n\ge 4)$ with nonnegative curvature operator and positive Ricci curvature away from a compact set of $M$. We prove that any $(n-1)$-dimensional compact split ancient solution from the blow-down of $(M, g)$ is of type I. The result is a generalization of our previous work from $n=4$ to any dimension.

Let $(M^n,g)$ $(n\ge 4)$ be a complete noncompact $\kappa$-noncollapsed steady Ricci soliton with $\rm{Rm}\geq 0$ and $\rm{Ric}> 0$ away from a compact set $K$ of $M$. We prove that there is no any $(n-1)$-dimensional compact split limit Ricci flow of type I arising from the blow-down of $(M, g)$, if there is an $(n-1)$-dimensional noncompact split limit Ricci flow. Consequently, the compact split limit ancient flows of type I and type II cannot occur simultaneously from the blow-down. As an application, we prove that $(M^n,g)$ with $\rm{Rm}\geq 0$ must be isometric the Bryant Ricci soliton up to scaling, if there exists a sequence of rescaled Ricci flows $(M,g_{p_i}(t); p_i)$ of $(M,g)$ converges subsequently to a family of shrinking quotient cylinders., Comment: arXiv admin note: text overlap with arXiv:2310.12529

In the paper, we analysis the asymptotic behavior of noncompact $\kappa$-noncollapsed steady gradient Ricci soliton $(M, g)$ with nonnegative curvature operator away from a compact set $K$ of $M$. In particular, we prove: any $4d$ noncompact $\kappa$-noncollapsed steady gradient Ricci soliton $(M^4, g)$ with nonnegative sectional curvature must be a Bryant Ricci soliton up to scaling if it admits a sequence of rescaled flows of $(M^4, g)$, which converges subsequently to a family of shrinking quotient cylinders., Comment: Proposition 4.1, Lemma 4.2 and Lemma 4.3 have been generalized for any dimension

We construct a doped holographic superconductor in the Gubser-Rocha model, and realize a superconducting dome in the middle of the temperature-doping phase diagram. It is worth noting that unlike the previous researches, the profile of our dome shrinks inward near zero temperature. From the numerical observation for the coupling dependence of the phase diagram, we find that the coupling between the two gauge fields plays a crucial role in the formation of dome. We also analytically calculate the DC conductivity of the normal phase of the system in the momentum dissipation and obtain the resistivity which is proportional to temperature. The AC conductivity is calculated numerically., Comment: 17 pages,10 figures

In this paper, we propose a technique for making humans in photographs protrude like reliefs. Unlike previous methods which mostly focus on the face and head, our method aims to generate art works that describe the whole body activity of the character. One challenge is that there is no ground-truth for supervised deep learning. We introduce a sigmoid variant function to manipulate gradients tactfully and train our neural networks by equipping with a loss function defined in gradient domain. The second challenge is that actual photographs often across different light conditions. We used image-based rendering technique to address this challenge and acquire rendering images and depth data under different lighting conditions. To make a clear division of labor in network modules, a two-scale architecture is proposed to create high-quality relief from a single photograph. Extensive experimental results on a variety of scenes show that our method is a highly effective solution for generating digital 2.5D artwork from photographs., Comment: 10 pages, 11 figures

We design and test a low-loss interface between superconducting 3-dimensional microwave cavities and 2-dimensional circuits, where the coupling rate is highly tunable. This interface seamlessly integrates a loop antenna and a Josephson junction-based coupling element. We demonstrate that the loss added by connecting this interface to the cavity is 1.28 kHz, corresponding to an inverse quality factor of $1/(4.5 \times 10^6)$. Furthermore, we show that the cavity's external coupling rate to a 50 $\Omega$ transmission line can be tuned from negligibly small to over 3 orders of magnitude larger than its internal loss rate in a characteristic time of 3.2 ns. This switching speed does not impose additional limits on the coupling rate because it is much faster than the coupling rate. Moreover, the coupler can be controlled by low frequency signals to avoid interference with microwave signals near the cavity or qubit frequencies. Finally, the coupling element introduces a 0.04 Hz/photon self-Kerr nonlinearity to the cavity, remaining linear in high photon number operations.

We introduce a new curvature-pinching condition, which is weaker than the positive sectional curvature or PIC1, and then we prove several rigidity results for the rotationally symmetric solutions of steady Ricci solitons, i.e., the Bryant Ricci solitons., Comment: Some parts in Section 1 are revised

With the introduction of ChatGPT, the public's perception of AI-generated content (AIGC) has begun to reshape. Artificial intelligence has significantly reduced the barrier to entry for non-professionals in creative endeavors, enhancing the efficiency of content creation. Recent advancements have seen significant improvements in the quality of symbolic music generation, which is enabled by the use of modern generative algorithms to extract patterns implicit in a piece of music based on rule constraints or a musical corpus. Nevertheless, existing literature reviews tend to present a conventional and conservative perspective on future development trajectories, with a notable absence of thorough benchmarking of generative models. This paper provides a survey and analysis of recent intelligent music generation techniques, outlining their respective characteristics and discussing existing methods for evaluation. Additionally, the paper compares the different characteristics of music generation techniques in the East and West as well as analysing the field's development prospects.

Unsupervised self-rehabilitation exercises and physical training can cause serious injuries if performed incorrectly. We introduce a learning-based framework that identifies the mistakes made by a user and proposes corrective measures for easier and safer individual training. Our framework does not rely on hard-coded, heuristic rules. Instead, it learns them from data, which facilitates its adaptation to specific user needs. To this end, we use a Graph Convolutional Network (GCN) architecture acting on the user's pose sequence to model the relationship between the body joints trajectories. To evaluate our approach, we introduce a dataset with 3 different physical exercises. Our approach yields 90.9% mistake identification accuracy and successfully corrects 94.2% of the mistakes., Comment: Video: https://youtu.be/W3kyyeHe0SI