arXiv subjects

Yanhui Huang

Publications and source records attributed to Yanhui Huang.

4 recordsLinked to original sources

arXiv2025

Compass-Thinker-7B Technical Report

Recent R1-Zero-like research further demonstrates that reasoning extension has given large language models (LLMs) unprecedented reasoning capabilities, and Reinforcement Learning is the core technology to elicit its complex reasoning. However, conducting RL experiments directly on hyperscale models involves high computational costs and resource demands, posing significant risks. We propose the Compass-Thinker-7B model, which aims to explore the potential of Reinforcement Learning with less computational resources and costs, and provides insights for further research into RL recipes for larger models. Compass-Thinker-7B is trained from an open source model through a specially designed Reinforcement Learning Pipeline. We curate a dataset of 30k verifiable mathematics problems for the Reinforcement Learning Pipeline. By configuring data and training settings with different difficulty distributions for different stages, the potential of the model is gradually released and the training efficiency is improved. Extensive evaluations show that Compass-Thinker-7B possesses exceptional reasoning potential, and achieves superior performance on mathematics compared to the same-sized RL model. Especially in the challenging AIME2024 evaluation, Compass-Thinker-7B achieves 40% accuracy.

cs.AI↗

arXiv2021

Learning Robust and Lightweight Model through Separable Structured Transformations

With the proliferation of mobile devices and the Internet of Things, deep learning models are increasingly deployed on devices with limited computing resources and memory, and are exposed to the threat of adversarial noise. Learning deep models with both lightweight and robustness is necessary for these equipments. However, current deep learning solutions are difficult to learn a model that possesses these two properties without degrading one or the other. As is well known, the fully-connected layers contribute most of the parameters of convolutional neural networks. We perform a separable structural transformation of the fully-connected layer to reduce the parameters, where the large-scale weight matrix of the fully-connected layer is decoupled by the tensor product of several separable small-sized matrices. Note that data, such as images, no longer need to be flattened before being fed to the fully-connected layer, retaining the valuable spatial geometric information of the data. Moreover, in order to further enhance both lightweight and robustness, we propose a joint constraint of sparsity and differentiable condition number, which is imposed on these separable matrices. We evaluate the proposed approach on MLP, VGG-16 and Vision Transformer. The experimental results on datasets such as ImageNet, SVHN, CIFAR-100 and CIFAR10 show that we successfully reduce the amount of network parameters by 90%, while the robust accuracy loss is less than 1.5%, which is better than the SOTA methods based on the original fully-connected layer. Interestingly, it can achieve an overwhelming advantage even at a high compression rate, e.g., 200 times.

cs.CV↗

arXiv2016

Understanding the Strain-Dependent Dielectric Behavior of Carbon Black Reinforced Natural Rubber: An interfacial or bulk phenomenon?

Filler-polymer interactions are one of the keys to understanding the physical properties of polymer composites. These interactions give rise to an interface with specific properties that may have a nontrivial effect on the macroscopic properties of composites. Direct measurement of the interface properties at nanometer scale is usually unavailable. Thus, interface properties are often back calculated from the bulk response using a computational model. However, if the model does not take into account the morphology of the filler dispersion, the results can be misleading. Recently it has been found that the dielectric response of a carbon black filled natural rubber film can change dramatically upon stretching [Huang, Macromolecules 49, 2339 (2016)]1. In this paper, we will show that this phenomenon can be largely explained by changes in filler cluster connectivity due to strain and is probably not caused by changes in the interfacial interactions. To support the argument, the polarization mechanism of the composites in the measured frequency range is analyzed and numerical models are developed to virtually reproduce the physical phenomenon as a function of strain. As a result, a power-law dependence of dielectric permittivity with strain is derived, which matches closely with the experimental results.

cond-mat.mtrl-sci↗

arXiv2016

On the Nature of High Field Charge Transport in Reinforced Silicone Dielectrics: Experiment and Simulation

The high field charge injection and transport properties in reinforced silicone dielectrics were investigated by measuring the time-dependent space charge distribution and the current under dc conditions up to the breakdown field, and were compared with properties of other dielectric polymers. It is argued that the energy and spatial distribution of localized electronic states are crucial to determining these properties for polymer dielectrics. Tunneling to localized states likely dominates the charge injection process. A transient transport regime arises due to the relaxation of charge carriers into deep traps at the energy band tails, and is successfully verified by a Monte Carlo simulation using the multiple-hopping model. The charge carrier mobility is found to be highly heterogeneous due to non-uniform trapping. The slow moving electron packet exhibits a negative field dependent drift velocity possibly due to the spatial disorder of traps.

cond-mat.mtrl-sci↗