Picture for Fangcheng Liu

Fangcheng Liu

Kangaroo: Lossless Self-Speculative Decoding via Double Early Exiting

Add code
Apr 29, 2024
Viaarxiv icon

Rethinking Optimization and Architecture for Tiny Language Models

Add code
Feb 06, 2024
Viaarxiv icon

PanGu-$π$: Enhancing Language Model Architectures via Nonlinearity Compensation

Add code
Dec 27, 2023
Figure 1 for PanGu-$π$: Enhancing Language Model Architectures via Nonlinearity Compensation
Figure 2 for PanGu-$π$: Enhancing Language Model Architectures via Nonlinearity Compensation
Figure 3 for PanGu-$π$: Enhancing Language Model Architectures via Nonlinearity Compensation
Figure 4 for PanGu-$π$: Enhancing Language Model Architectures via Nonlinearity Compensation
Viaarxiv icon

Towards Transferable Unrestricted Adversarial Examples with Minimum Changes

Add code
Jan 04, 2022
Figure 1 for Towards Transferable Unrestricted Adversarial Examples with Minimum Changes
Figure 2 for Towards Transferable Unrestricted Adversarial Examples with Minimum Changes
Figure 3 for Towards Transferable Unrestricted Adversarial Examples with Minimum Changes
Figure 4 for Towards Transferable Unrestricted Adversarial Examples with Minimum Changes
Viaarxiv icon

Unrestricted Adversarial Attacks on ImageNet Competition

Add code
Oct 25, 2021
Figure 1 for Unrestricted Adversarial Attacks on ImageNet Competition
Figure 2 for Unrestricted Adversarial Attacks on ImageNet Competition
Figure 3 for Unrestricted Adversarial Attacks on ImageNet Competition
Viaarxiv icon