Position: TrustLLM: Trustworthiness in Large Language Models

Yue Huang, Lichao Sun, Haoran Wang, Siyuan Wu, Qihui Zhang, Yuan Li, Chujie Gao, Yixin Huang, Wenhan Lyu, Yixuan Zhang, Xiner Li, Hanchi Sun, Zhengliang Liu, Yixin Liu, Yijue Wang, Zhikun Zhang, Bertie Vidgen, Bhavya Kailkhura, Caiming Xiong, Chaowei Xiao, Chunyuan Li, Eric P. Xing, Furong Huang, Hao Liu, Heng Ji, Hongyi Wang, Huan Zhang, Huaxiu Yao, Manolis Kellis, Marinka Zitnik, Meng Jiang, Mohit Bansal, James Zou, Jian Pei, Jian Liu, Jianfeng Gao, Jiawei Han, Jieyu Zhao, Jiliang Tang, Jindong Wang, Joaquin Vanschoren, John Mitchell, Kai Shu, Kaidi Xu, Kai-Wei Chang, Lifang He, Lifu Huang, Michael Backes, Neil Zhenqiang Gong, Philip S. Yu, Pin-Yu Chen, Quanquan Gu, Ran Xu, Rex Ying, Shuiwang Ji, Suman Jana, Tianlong Chen, Tianming Liu, Tianyi Zhou, William Yang Wang, Xiang Li, Xiangliang Zhang, Xiao Wang, Xing Xie, Xun Chen, Xuyu Wang, Yan Liu, Yanfang Ye, Yinzhi Cao, Yong Chen, Yue Zhao
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:20166-20270, 2024.

Abstract

Large language models (LLMs) have gained considerable attention for their excellent natural language processing capabilities. Nonetheless, these LLMs present many challenges, particularly in the realm of trustworthiness. This paper introduces TrustLLM, a comprehensive study of trustworthiness in LLMs, including principles for different dimensions of trustworthiness, established benchmark, evaluation, and analysis of trustworthiness for mainstream LLMs, and discussion of open challenges and future directions. Specifically, we first propose a set of principles for trustworthy LLMs that span eight different dimensions. Based on these principles, we further establish a benchmark across six dimensions including truthfulness, safety, fairness, robustness, privacy, and machine ethics. We then present a study evaluating 16 mainstream LLMs in TrustLLM, consisting of over 30 datasets. Our findings firstly show that in general trustworthiness and capability (i.e., functional effectiveness) are positively related. Secondly, our observations reveal that proprietary LLMs generally outperform most open-source counterparts in terms of trustworthiness, raising concerns about the potential risks of widely accessible open-source LLMs. However, a few open-source LLMs come very close to proprietary ones, suggesting that open-source models can achieve high levels of trustworthiness without additional mechanisms like moderator, offering valuable insights for developers in this field. Thirdly, it is important to note that some LLMs may be overly calibrated towards exhibiting trustworthiness, to the extent that they compromise their utility by mistakenly treating benign prompts as harmful and consequently not responding. Besides these observations, we’ve uncovered key insights into the multifaceted trustworthiness in LLMs. We emphasize the importance of ensuring transparency not only in the models themselves but also in the technologies that underpin trustworthiness. We advocate that the establishment of an AI alliance between industry, academia, the open-source community to foster collaboration is imperative to advance the trustworthiness of LLMs.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-huang24x, title = {Position: {T}rust{LLM}: Trustworthiness in Large Language Models}, author = {Huang, Yue and Sun, Lichao and Wang, Haoran and Wu, Siyuan and Zhang, Qihui and Li, Yuan and Gao, Chujie and Huang, Yixin and Lyu, Wenhan and Zhang, Yixuan and Li, Xiner and Sun, Hanchi and Liu, Zhengliang and Liu, Yixin and Wang, Yijue and Zhang, Zhikun and Vidgen, Bertie and Kailkhura, Bhavya and Xiong, Caiming and Xiao, Chaowei and Li, Chunyuan and Xing, Eric P. and Huang, Furong and Liu, Hao and Ji, Heng and Wang, Hongyi and Zhang, Huan and Yao, Huaxiu and Kellis, Manolis and Zitnik, Marinka and Jiang, Meng and Bansal, Mohit and Zou, James and Pei, Jian and Liu, Jian and Gao, Jianfeng and Han, Jiawei and Zhao, Jieyu and Tang, Jiliang and Wang, Jindong and Vanschoren, Joaquin and Mitchell, John and Shu, Kai and Xu, Kaidi and Chang, Kai-Wei and He, Lifang and Huang, Lifu and Backes, Michael and Gong, Neil Zhenqiang and Yu, Philip S. and Chen, Pin-Yu and Gu, Quanquan and Xu, Ran and Ying, Rex and Ji, Shuiwang and Jana, Suman and Chen, Tianlong and Liu, Tianming and Zhou, Tianyi and Wang, William Yang and Li, Xiang and Zhang, Xiangliang and Wang, Xiao and Xie, Xing and Chen, Xun and Wang, Xuyu and Liu, Yan and Ye, Yanfang and Cao, Yinzhi and Chen, Yong and Zhao, Yue}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {20166--20270}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/huang24x/huang24x.pdf}, url = {https://proceedings.mlr.press/v235/huang24x.html}, abstract = {Large language models (LLMs) have gained considerable attention for their excellent natural language processing capabilities. Nonetheless, these LLMs present many challenges, particularly in the realm of trustworthiness. This paper introduces TrustLLM, a comprehensive study of trustworthiness in LLMs, including principles for different dimensions of trustworthiness, established benchmark, evaluation, and analysis of trustworthiness for mainstream LLMs, and discussion of open challenges and future directions. Specifically, we first propose a set of principles for trustworthy LLMs that span eight different dimensions. Based on these principles, we further establish a benchmark across six dimensions including truthfulness, safety, fairness, robustness, privacy, and machine ethics. We then present a study evaluating 16 mainstream LLMs in TrustLLM, consisting of over 30 datasets. Our findings firstly show that in general trustworthiness and capability (i.e., functional effectiveness) are positively related. Secondly, our observations reveal that proprietary LLMs generally outperform most open-source counterparts in terms of trustworthiness, raising concerns about the potential risks of widely accessible open-source LLMs. However, a few open-source LLMs come very close to proprietary ones, suggesting that open-source models can achieve high levels of trustworthiness without additional mechanisms like moderator, offering valuable insights for developers in this field. Thirdly, it is important to note that some LLMs may be overly calibrated towards exhibiting trustworthiness, to the extent that they compromise their utility by mistakenly treating benign prompts as harmful and consequently not responding. Besides these observations, we’ve uncovered key insights into the multifaceted trustworthiness in LLMs. We emphasize the importance of ensuring transparency not only in the models themselves but also in the technologies that underpin trustworthiness. We advocate that the establishment of an AI alliance between industry, academia, the open-source community to foster collaboration is imperative to advance the trustworthiness of LLMs.} }
Endnote
%0 Conference Paper %T Position: TrustLLM: Trustworthiness in Large Language Models %A Yue Huang %A Lichao Sun %A Haoran Wang %A Siyuan Wu %A Qihui Zhang %A Yuan Li %A Chujie Gao %A Yixin Huang %A Wenhan Lyu %A Yixuan Zhang %A Xiner Li %A Hanchi Sun %A Zhengliang Liu %A Yixin Liu %A Yijue Wang %A Zhikun Zhang %A Bertie Vidgen %A Bhavya Kailkhura %A Caiming Xiong %A Chaowei Xiao %A Chunyuan Li %A Eric P. Xing %A Furong Huang %A Hao Liu %A Heng Ji %A Hongyi Wang %A Huan Zhang %A Huaxiu Yao %A Manolis Kellis %A Marinka Zitnik %A Meng Jiang %A Mohit Bansal %A James Zou %A Jian Pei %A Jian Liu %A Jianfeng Gao %A Jiawei Han %A Jieyu Zhao %A Jiliang Tang %A Jindong Wang %A Joaquin Vanschoren %A John Mitchell %A Kai Shu %A Kaidi Xu %A Kai-Wei Chang %A Lifang He %A Lifu Huang %A Michael Backes %A Neil Zhenqiang Gong %A Philip S. Yu %A Pin-Yu Chen %A Quanquan Gu %A Ran Xu %A Rex Ying %A Shuiwang Ji %A Suman Jana %A Tianlong Chen %A Tianming Liu %A Tianyi Zhou %A William Yang Wang %A Xiang Li %A Xiangliang Zhang %A Xiao Wang %A Xing Xie %A Xun Chen %A Xuyu Wang %A Yan Liu %A Yanfang Ye %A Yinzhi Cao %A Yong Chen %A Yue Zhao %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-huang24x %I PMLR %P 20166--20270 %U https://proceedings.mlr.press/v235/huang24x.html %V 235 %X Large language models (LLMs) have gained considerable attention for their excellent natural language processing capabilities. Nonetheless, these LLMs present many challenges, particularly in the realm of trustworthiness. This paper introduces TrustLLM, a comprehensive study of trustworthiness in LLMs, including principles for different dimensions of trustworthiness, established benchmark, evaluation, and analysis of trustworthiness for mainstream LLMs, and discussion of open challenges and future directions. Specifically, we first propose a set of principles for trustworthy LLMs that span eight different dimensions. Based on these principles, we further establish a benchmark across six dimensions including truthfulness, safety, fairness, robustness, privacy, and machine ethics. We then present a study evaluating 16 mainstream LLMs in TrustLLM, consisting of over 30 datasets. Our findings firstly show that in general trustworthiness and capability (i.e., functional effectiveness) are positively related. Secondly, our observations reveal that proprietary LLMs generally outperform most open-source counterparts in terms of trustworthiness, raising concerns about the potential risks of widely accessible open-source LLMs. However, a few open-source LLMs come very close to proprietary ones, suggesting that open-source models can achieve high levels of trustworthiness without additional mechanisms like moderator, offering valuable insights for developers in this field. Thirdly, it is important to note that some LLMs may be overly calibrated towards exhibiting trustworthiness, to the extent that they compromise their utility by mistakenly treating benign prompts as harmful and consequently not responding. Besides these observations, we’ve uncovered key insights into the multifaceted trustworthiness in LLMs. We emphasize the importance of ensuring transparency not only in the models themselves but also in the technologies that underpin trustworthiness. We advocate that the establishment of an AI alliance between industry, academia, the open-source community to foster collaboration is imperative to advance the trustworthiness of LLMs.
APA
Huang, Y., Sun, L., Wang, H., Wu, S., Zhang, Q., Li, Y., Gao, C., Huang, Y., Lyu, W., Zhang, Y., Li, X., Sun, H., Liu, Z., Liu, Y., Wang, Y., Zhang, Z., Vidgen, B., Kailkhura, B., Xiong, C., Xiao, C., Li, C., Xing, E.P., Huang, F., Liu, H., Ji, H., Wang, H., Zhang, H., Yao, H., Kellis, M., Zitnik, M., Jiang, M., Bansal, M., Zou, J., Pei, J., Liu, J., Gao, J., Han, J., Zhao, J., Tang, J., Wang, J., Vanschoren, J., Mitchell, J., Shu, K., Xu, K., Chang, K., He, L., Huang, L., Backes, M., Gong, N.Z., Yu, P.S., Chen, P., Gu, Q., Xu, R., Ying, R., Ji, S., Jana, S., Chen, T., Liu, T., Zhou, T., Wang, W.Y., Li, X., Zhang, X., Wang, X., Xie, X., Chen, X., Wang, X., Liu, Y., Ye, Y., Cao, Y., Chen, Y. & Zhao, Y.. (2024). Position: TrustLLM: Trustworthiness in Large Language Models. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:20166-20270 Available from https://proceedings.mlr.press/v235/huang24x.html.

Related Material