Zobrazeno 1 - 10
of 35
pro vyhledávání: '"Wu, Xianchao"'
Autor:
Xu, Peng, Ping, Wei, Wu, Xianchao, Xu, Chejian, Liu, Zihan, Shoeybi, Mohammad, Catanzaro, Bryan
In this work, we introduce ChatQA 2, an Llama 3.0-based model with a 128K context window, designed to bridge the gap between open-source LLMs and leading proprietary models (e.g., GPT-4-Turbo) in long-context understanding and retrieval-augmented gen
Externí odkaz:
http://arxiv.org/abs/2407.14482
Autor:
Wu, Xianchao, Zhang, Lan
The problem of obtaining the lower bounds on the restriction of Laplacian eigenfunctions to hypersurfaces inside a compact Riemannian manifold $(M,g)$ is challenging and has been attempted by many authors \cite{BR, GRS, Jun, ET}. This paper aims to s
Externí odkaz:
http://arxiv.org/abs/2403.19188
Autor:
Wu, Xianchao
Let $\{u_\lambda\}$ be a sequence of $L^2$-normalized Laplacian eigenfunctions on a compact two-dimensional smooth Riemanniann manifold $(M,g)$. We seek to get an $L^p$ restriction bounds of the Neumann data $ \lambda^{-1} \partial_\nu u_{\lambda}\,\
Externí odkaz:
http://arxiv.org/abs/2403.16445
Model alignment with human preferences is an essential step in making Large Language Models (LLMs) helpful and consistent with human values. It typically consists of supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF) s
Externí odkaz:
http://arxiv.org/abs/2310.05344
Autor:
Xu, Peng, Ping, Wei, Wu, Xianchao, McAfee, Lawrence, Zhu, Chen, Liu, Zihan, Subramanian, Sandeep, Bakhturina, Evelina, Shoeybi, Mohammad, Catanzaro, Bryan
Extending the context window of large language models (LLMs) is getting popular recently, while the solution of augmenting LLMs with retrieval has existed for years. The natural questions are: i) Retrieval-augmentation versus long context window, whi
Externí odkaz:
http://arxiv.org/abs/2310.03025
Autor:
Wu, Xianchao
Speech-to-speech translation is a typical sequence-to-sequence learning task that naturally has two directions. How to effectively leverage bidirectional supervision signals to produce high-fidelity audio for both directions? Existing approaches eith
Externí odkaz:
http://arxiv.org/abs/2305.12628
Autor:
Wu, Xianchao
We enhance the vanilla adversarial training method for unsupervised Automatic Speech Recognition (ASR) by a diffusion-GAN. Our model (1) injects instance noises of various intensities to the generator's output and unlabeled reference text which are s
Externí odkaz:
http://arxiv.org/abs/2303.13559
Autor:
Wu, Xianchao
Artistic painting has achieved significant progress during recent years. Using an autoencoder to connect the original images with compressed latent spaces and a cross attention enhanced U-Net as the backbone of diffusion, latent diffusion models (LDM
Externí odkaz:
http://arxiv.org/abs/2209.14697
Autor:
Wu, Xianchao
Citrinet is an end-to-end convolutional Connectionist Temporal Classification (CTC) based automatic speech recognition (ASR) model. To capture local and global contextual information, 1D time-channel separable convolutions combined with sub-word enco
Externí odkaz:
http://arxiv.org/abs/2209.00261
Autor:
Wu, Xianchao
Conformer has achieved impressive results in Automatic Speech Recognition (ASR) by leveraging transformer's capturing of content-based global interactions and convolutional neural network's exploiting of local features. In Conformer, two macaron-like
Externí odkaz:
http://arxiv.org/abs/2209.00260