This page is still being polished. If you have thoughts, please share them via the feedback form.
Data on this page is preliminary and may change. Please do not share or cite these figures publicly.
Techniques to remove, bound, or modify learned model capabilities post-training.
Also in Model
Interpretability can identify where and how facts are stored in LMs, how they are recalled during reasoning, and how to direct activations towards facts through knowledge editing methods, thereby avoiding hallucinations
Meng et al. (2022) use path patching to locate the components responsible for storing factual knowledge, and then edit the facts by updating only the parameters of these components (e.g., replacing “Seattle” with “Paris”). Another study investigates the sources of hallucinations through perturbation analysis of source token contribution patterns (Xu et al., 2023c). Their findings suggest that hallucinations may stem from the model’s over-reliance on a limited set of source tokens. Duan et al. (2024a) apply PCA to derive the direction of the final hidden state corresponding to the correct answer and used this direction to enhance the hidden representation to reduce hallucinations. Although these methods are effective for targeted editing, their ability to update relevant knowledge and prevent forgetting still requires further research (Cohen et al., 2023).
Reasoning
Hallucination mitigation spans multiple model-level approaches without clear focal activity specification.
Value Misalignment
99.9 OtherValue Misalignment > Mitigating social bias
1 AI SystemValue Misalignment > Privacy protection
1 AI SystemValue Misalignment > Methods for mitigating toxicity
1 AI SystemValue Misalignment > Methods for mitigating LLM amorality
1 AI SystemRobustness to attack
1 AI SystemLarge Language Model Safety: A Holistic Survey
Shi, Dan; Shen, Tianhao; Huang, Yufei; Li, Zhigen; Leng, Yongqi; Jin, Renren; Liu, Chuang; Wu, Xinwei; Guo, Zishan; Yu, Linhao; Shi, Ling; Jiang, Bojian; Xiong, Deyi (2024)
The rapid development and deployment of large language models (LLMs) have introduced a new frontier in artificial intelligence, marked by unprecedented capabilities in natural language understanding and generation. However, the increasing integration of these models into critical applications raises substantial safety concerns, necessitating a thorough examination of their potential risks and associated mitigation strategies. This survey provides a comprehensive overview of the current landscape of LLM safety, covering four major categories: value misalignment, robustness to adversarial attacks, misuse, and autonomous AI risks. In addition to the comprehensive review of the mitigation methodologies and evaluation resources on these four aspects, we further explore four topics related to LLM safety: the safety implications of LLM agents, the role of interpretability in enhancing LLM safety, the technology roadmaps proposed and abided by a list of AI companies and institutes for LLM safety, and AI governance aimed at LLM safety with discussions on international cooperation, policy proposals, and prospective regulatory directions. Our findings underscore the necessity for a proactive, multifaceted approach to LLM safety, emphasizing the integration of technical solutions, ethical considerations, and robust governance frameworks. This survey is intended to serve as a foundational resource for academy researchers, industry practitioners, and policymakers, offering insights into the challenges and opportunities associated with the safe integration of LLMs into society. Ultimately, it seeks to contribute to the safe and beneficial development of LLMs, aligning with the overarching goal of harnessing AI for societal advancement and well-being. A curated list of related papers has been publicly available at https://github.com/tjunlp-lab/Awesome-LLM-Safety-Papers.
Build and Use Model
Training, fine-tuning, and integrating the AI model
Developer
Entity that creates, trains, or modifies the AI system
Manage
Prioritising, responding to, and mitigating AI risks
