Name: chrisliu298
Type: User
Company: University of California, Santa Cruz
Bio: PhD student @UCSC CSE | Researcher at @SkyworkAI
Twitter: chrisliu298
Location: Santa Cruz, California
Blog: chrisliu298.ai
chrisliu298's Projects
We show that adversarially training (Fast Gradient Sign Method and Projected Gradient Descent) reduces the empirically sample complexity rate for MLP and a variety of CNN architectures on MNIST and CIFAR-10.
Robust recipes to align language models with human and AI preferences
I use the titles and abstract of these papers to fine-tune my GPT-2 model.
A resource repository for machine unlearning in large language models
A resource repository for representation engineering in large language models
A resource repository of sparse autoencoders for large language models
I train several models as baseline on CIFAR-10 and CIFAR-100.
Improving Alignment and Robustness with Circuit Breakers
Learning to construct compact representations of neural network weight matrices by linearly projecting them into a smaller space
Neovim plugin for GitHub Copilot
Academic CVs that you can emulate
Reference implementation for DPO (Direct Preference Optimization)
Oneline, minimal, keyboard-centered Firefox CSS theme.
Fine-tuning GPT-2 to generate research paper abstracts
Minimalist Hugo template for academic websites
Deep Learning for humans
A launch point for your personal nvim configuration
Code for Large Language Model Unlearning via Embedding-Corrupted Prompts