Leaked GPTs Prompts Bypass the 25 message limit or to try out GPTs without a Plus subscription.

Code of NAACL 2024 paper "Stealthy and Persistent Unalignment on Large Language Models via Backdoor Injections".

A simple pip-installable Python tool to generate your own HTML citation world map from your Google Scholar ID.

Open source real-time translation app for Android that runs locally

This is the source code for Data-free Backdoor. Our paper is accepted by the 32nd USENIX Security Symposium (USENIX Security 2023).

[ICML 2024] Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank Modifications

VideoSys: An easy and efficient system for video generation

Refine high-quality datasets and visual AI models

Code for MBGE-recognition: Emotion recognition based on multi-view body gestures, accepted at ICIP 2019.

Code for LGRe: Latent Graph Recurrent Network for Document Ranking, accepted at DASFAA 2021.

Code for R-former: Legal Judgment Prediction via Relational Learning, accepted at SIGIR 2021.

Code for KERM: Incorporating Explicit Knowledge in Pre-trained Language Models for Passage Re-ranking, accepted at SIGIR 2022.

code of paper "Defending Against Alignment-Breaking Attacks via Robustly Aligned LLM"

code of paper "IMPRESS: Evaluating the Resilience of Imperceptible Perturbations Against Unauthorized Data Usage in Diffusion-Based Generative AI"

Can large language models provide useful feedback on research papers? A large-scale empirical analysis.

抢占显卡

A framework to evaluate the generalization capability of safety alignment for LLMs

The automated prompt injection framework for LLM-integrated applications.

Universal and Transferable Attacks on Aligned Language Models

Awesome-LLM-Robustness: a curated list of Uncertainty, Reliability and Robustness in Large Language Models

Build agents which are controlled by LLMs

🦜🔗 Build context-aware reasoning applications

All image quality metrics you need in one package.

Code for loralib, an implementation of "LoRA: Low-Rank Adaptation of Large Language Models"

Using Low-rank adaptation to quickly fine-tune diffusion models.

Let us control diffusion models!

Hackable and optimized Transformers building blocks, supporting a composable construction.

Raising the Cost of Malicious AI-Powered Image Editing

A new adversarial purification method that uses the forward and reverse processes of diffusion models to remove adversarial perturbations.

Benchmark for Natural Temporal Distribution Shift (NeurIPS 2022)