Awesome-LM-SSP: A Deep Dive into Trustworthy Large Models
Introduction
The Awesome-LM-SSP project is a comprehensive resource that focuses on the trustworthiness of large models (LMs) across several dimensions, specifically concentrating on multi-modal LMs such as vision-language models and diffusion models. The project aims to address key concerns related to the safety, security, and privacy of these large models.
This initiative is an ongoing project, and it’s gradually being enriched with valuable content. It stands as an essential collection for anyone interested in understanding and improving the trustworthiness of large models in various applications.
Project Badges and Classification
Various icons or badges are used in the project to categorize and represent different aspects, including:
-
Model Types:
- Large Language Model (LLM)
- Vision Language Model (VLM)
- Speech Language Model (SLM)
- Diffusion Models
-
Content Tags:
- Benchmark
- New Dataset
- Agent
- Code Generation
- Defense
- Retrieval-Augmented Generation (RAG)
- Chinese-specific resources
-
Venues:
- Conference
- Blog
- OpenAI
- Meta AI
The project encourages contributions from the community and welcomes suggestions for new resources.
Latest News
The project keeps its audience updated with the latest developments in the field:
- A significant number of research papers have been collected from major conferences like ACL, S&P, NDSS, and ICLR in 2024.
- The LM-SSP was officially released in January 2024, marking its proactive engagement in the field.
Collected Resources
The project hosts an extensive collection of resources categorized into various topics. Notable sections include:
- Books and Competitions
- Leaderboards and Toolkits
- Surveys
The paper collections are particularly detailed:
- Safety: Covering topics from general safety to specific issues like jailbreak, ethical concerns, fairness, and hallucination.
- Security: Encompassing adversarial examples and strategies against potential backdoor attacks.
- Privacy: Addressing areas like data contamination, copyright issues, and privacy-preserving computations.
Star History
The project provides a visual representation of its popularity and engagement over time, showcasing its growth and community interest.
Acknowledgement
The project is coordinated by a group of dedicated organizers including Tianshuo Cong, Xinlei He, Zhengyu Zhao, Yugeng Liu, and Delong Ran. It draws inspiration from several existing projects like LLM Security and PLMpapers, aiming to build upon and expand the knowledge repository for trustworthy LMs.
As a dynamic and evolving project, Awesome-LM-SSP is a cornerstone for researchers, developers, and anyone vested in the future of artificial intelligence, ensuring models remain trustworthy and secure across various applications.