The official repository of ECCV 2024 paper "Outlier-Aware Test-time Adaptation with Stable Memory Replay"

Official PyTorch implementation of Learning to (Learn at Test Time): RNNs with Expressive Hidden States

[知识编辑] Must-read Papers on Knowledge Editing for Large Language Models.

SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal Behaviors

code for our ICML-2024 paper "Realistic Unsupervised CLIP Fine-tuning with Universal Entropy Optimization"

Official code for ICML 2024 paper, "Connecting the Dots: Collaborative Fine-tuning for Black-Box Vision-Language Models"

MambaOut: Do We Really Need Mamba for Vision?

WMDP is a LLM proxy benchmark for hazardous knowledge in bio, cyber, and chemical security. We also release code for RMU, an unlearning method which reduces LLM performance on WMDP while retaining …

A curated list of papers & resources linked to data poisoning, backdoor attacks and defenses against them

[arXiv2024] Unsolvable Problem Detection: Evaluating Trustworthiness of Vision Language Models

Tools for merging pretrained large language models.

The official repo for the paper "VeCLIP: Improving CLIP Training via Visual-enriched Captions"

[NAACL2024] Attacks, Defenses and Evaluations for LLM Conversation Safety: A Survey

A collection of model transferability estimation methods.

A comprehensive toolbox for model inversion attacks and defenses, which is easy to get started.

HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal

Code base of SynthCLIP: CLIP training with purely synthetic text-image pairs from LLMs and TTIs.

[ICLR 2024] Towards Elminating Hard Label Constraints in Gradient Inverision Attacks

Fingerprint large language models

Official code for ICLR 2024 paper, "A Hard-to-Beat Baseline for Training-free CLIP-based Adaptation"

A reading list for large models safety, security, and privacy (including Awesome LLM Security, Safety, etc.).