IBM 科技百花 2023-07-24 15:19:51 Multi-Symmetry Ensembles: Improving Diversity and Generalization via Opposing Symmetries Download paperAbstractDeep ensembles (DE) have been successful in improving model performance by lea... 评论 分享 微信扫一扫：分享 微信点“扫一扫” 点右上“...”，便可分享本文
IBM nightclub 2023-07-24 13:43:18 PAC Generalization via Invariant Representations Download paperAbstractInvariant representations are transformations of the covariates such that the ... 评论 分享 微信扫一扫：分享 微信点“扫一扫” 点右上“...”，便可分享本文
IBM nightclub 2023-07-24 13:32:40 PromptBoosting: Black-Box Text Classification with Ten Forward Passes Download paperAbstractWe describe PromptBoosting, a query-efficient procedure for building a text cl... 评论 分享 微信扫一扫：分享 微信点“扫一扫” 点右上“...”，便可分享本文
IBM nightclub 2023-07-24 13:05:13 Graph Switching Dynamical Systems Download paperAbstractDynamical systems with complex behaviours, e.g. immune system cells interactin... 评论 分享 微信扫一扫：分享 微信点“扫一扫” 点右上“...”，便可分享本文
IBM 科技百花 2023-07-24 11:47:45 Data Efficient Neural Scaling Law via Model Reusing Download paperAbstractThe number of parameters in large transformers has been observed to grow expon... 评论 分享 微信扫一扫：分享 微信点“扫一扫” 点右上“...”，便可分享本文
IBM 大话互联网 2023-07-24 11:46:18 On the Forward Invariance of Neural ODEs Download paperAbstractWe propose a new method to ensure neural ordinary differential equations (ODEs... 评论 分享 微信扫一扫：分享 微信点“扫一扫” 点右上“...”，便可分享本文
IBM 科技百花 2023-07-24 10:00:52 Reparameterized Policy Learning for Multimodal Trajectory Optimization Download paperAbstractWe investigate the challenge of parametrizing policies for reinforcement learn... 评论 分享 微信扫一扫：分享 微信点“扫一扫” 点右上“...”，便可分享本文
IBM 大话互联网 2023-07-24 09:53:10 Learning Neural Constitutive Laws from Motion Observations for Generalizable PDE Dynamics Download paperAbstractWe propose a hybrid neural network (NN) and PDE approach for learning generali... 评论 分享 微信扫一扫：分享 微信点“扫一扫” 点右上“...”，便可分享本文
Meta nightclub 2023-07-18 11:43:00 Llama 2: Open Foundation and Fine-Tuned Chat Models AbstractIn this work, we develop and release Llama 2, a collection of pretrained and fine-tuned larg... 评论 分享 微信扫一扫：分享 微信点“扫一扫” 点右上“...”，便可分享本文
Meta 科技百花 2023-07-14 15:20:49 Scaling Autoregressive Multi-Modal Models: Pretraining and Instruction Tuning AbstractWe present CM3Leon (pronounced “Chameleon”), a retrieval-augmented, tokenbased, decoder-only... 评论 分享 微信扫一扫：分享 微信点“扫一扫” 点右上“...”，便可分享本文
IBM 科技百花 2023-07-09 12:45:52 AccShield: a New Trusted Execution Environment with Machine-Learning Accelerators View publicationAbstractMachine learning accelerators such as the Tensor Processing Unit (TPU) are a... 评论 分享 微信扫一扫：分享 微信点“扫一扫” 点右上“...”，便可分享本文
IBM nightclub 2023-07-09 09:07:18 DisentQA: Disentangling Parametric and Contextual Knowledge with Counterfactual Question Answering Download paperAbstractQuestion answering models commonly have access to two sources of "knowledge" d... 评论 分享 微信扫一扫：分享 微信点“扫一扫” 点右上“...”，便可分享本文
