Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignment

By Yang Liu, Yuanshun Yao, Jean-François Ton, Xiaoying Zhang, Ruocheng Guo, Hao Cheng, Yegor Klochkov, Muhammad Faaiz Taufiq, Hanguang Li. arXiv.org, 2023

575 citations recorded by the source corpus.

Ensuring alignment, which refers to making models behave in accordance with human intentions [1,2], has become a critical task before deploying large language models (LLMs) in real-world applications. For instance, OpenAI devoted six months to iteratively aligning GPT-4 before its release [3]. However, a major challenge faced by practitioners is the lack of clear guidance on evaluating whether LLM outputs align with social norms, values, and regulations. This obstacle hinders systematic iteration and deployment of LLMs. To address this issue, this paper presents a comprehensive survey of key dimensions that are crucial to consider when assessing LLM trustworthiness. The survey covers seven major categories of LLM trustworthiness: reliability, safety, fairness, resistance to misuse, explainability and reasoning, adherence to social norms, and robustness. Each major category is further divided into several sub-categories, resulting in a total of 29 sub-categories. Additionally, a subset of 8 sub-categories is selected for further investigation, where corresponding measurement studies are designed and conducted on several widely-used LLMs. The measurement results indicate that, in general, more aligned models tend to perform better in terms of overall trustworthiness. However, the effectiveness of alignment varies across the different trustworthiness categories considered. This highlights the importance of conducting more fine-grained analyses, testing, and making continuous improvements on LLM alignment. By shedding light on these key dimensions of LLM trustworthiness, this paper aims to provide valuable insights and guidance to practitioners in the field. Understanding and addressing these concerns will be crucial in achieving reliable and ethically sound deployment of LLMs in various applications.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/7142e920b6b9355d9cbacc9450818f912eca138e

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Xinwei Wu
2. Tianhao Shen
3. Weilong Dong
4. Chuang Liu
5. Renren Jin
6. Zishan Guo
7. Yan Liu
8. Deyi Xiong

Citations in the corpus

• Justice or Prejudice? Quantifying Biases in LLM-as-a-Judge (2024, background): “The imbalance in pre-training data can lead to imbalances during model training (Liu et al., 2024), resulting in biases against certain demographic groups, such as different genders (Wan et al., 2023), ages (Macnicol, 2006), and various languages (Jiao et al., 2023; Bang et al., 2023).”
• A Survey on the Memory Mechanism of Large Language Model-based Agents (2024, unknown): “In addition, cognitive psychology has been studied for a long time, so many effective human memory theories and architectures have been accumulated, which can support more advanced capabilities of the agents [147].”
• Security and Privacy Challenges of Large Language Models: A Survey (2024, unknown): “This involves reviewing existing studies in the broad area of security, privacy, LLMs, and their intersections, and notably, highlighting future research directions to design novel evaluation protocols and attack defense mechanisms tailored to the evolving landscape of LLMs.”
• Large Language Model Alignment: A Survey (2023, background): “In response to the ever-growing interest in this area, a few articles have recently reviewed (or incidentally discussed) alignment methods for LLMs (Pan et al., 2023; Zhao et al., 2023b; Fernandes et al., 2023; Liu et al., 2023d; Wang et al., 2023d).”
• Why Do Multi-Agent LLM Systems Fail? (2025, unknown): “These evaluations are crucial but primarily facilitate a top-down perspective, focusing on aggregate performance or high-level objectives like trustworthiness and security (Liu et al., 2023c; Yao et al., 2024b).”
• A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions (2023, unknown): “(Liu et al., 2023h) discusses the trustworthiness of LLMs from a broader perspective, while (Wang et al., 2023c) delves deeply into LLM factuality.”
• High-Dimension Human Value Representation in Large Language Models (2024, unknown): “Aligning LLMs with humans to enhance their service and mitigate risks has become a major focus [33].”