Large Language Models (LLMs) are highly sensitive to prompts, including additional context provided therein. As LLMs grow in capability, understanding their prompt-sensitivity becomes increasingly crucial for ensuring reliable and robust performance, particularly since evaluating these models becomes more challenging. In this work, we investigate how current models (Llama, Mixtral, Falcon) respond when presented with additional input from another model, mimicking a scenario where a more capable model -- or a system with access to more external information -- provides supplementary information to the target model. Across a diverse spectrum of question-answering tasks, we study how an LLM's response to multiple-choice questions changes when the prompt includes a prediction and explanation from another model. Specifically, we explore the influence of the presence of an explanation, the stated authoritativeness of the source, and the stated confidence of the supplementary input. Our findings reveal that models are strongly influenced, and when explanations are provided they are swayed irrespective of the quality of the explanation. The models are more likely to be swayed if the input is presented as being authoritative or confident, but the effect is small in size. This study underscores the significant prompt-sensitivity of LLMs and highlights the potential risks of incorporating outputs from external sources without thorough scrutiny and further validation. As LLMs continue to advance, understanding and mitigating such sensitivities will be crucial for their reliable and trustworthy deployment.
University of Michigan(密歇根大学);University Pompeu Fabra(庞培法布拉大学);Federal University of Minas Gerais(米纳斯吉拉斯联邦大学);IBM Research(IBM研究院);MIT-IBM Watson AI Lab(麻省理工学院-IBM沃森人工智能实验室);MIT(麻省理工学院)
Soft Begging: Modular and Efficient Shielding of LLMs against Prompt Injection and Jailbreaking based on Prompt Tuning
Deutsches Forschungszentrum für Künstliche Intelligenz (DFKI)(德国人工智能研究中心);Centre for European Research in Trusted AI (CERTAIN)(欧洲可信人工智能研究中心);sequire technology GmbH(sequire技术有限公司)
LLMs can persuade only psychologically susceptible humans on societal issues, via trust in AI and emotional appeals, amid logical fallacies
organization= CogNosco Lab, Department of Psychology;Cognitive Science, University of Trento , city= Rovereto , country= Italy;organization= Institute of Information Science;Technologies ``Alessandro Faedo", National Research Council , country= Italy;organization= Thomas Lord Department of Computer Science, University of Southern California , city= Los Angeles , state= California , country= USA;organization= Department of Sociology;Social Research, University of Trento , city= Trento , country= Italy;organization= Centre for Behavioural;Implementation Sciences in Medicine (BISI), National University of Singapore , city= Singapore , country= Singapore
Jailbreak Instruction-Tuned LLMs via end-of-sentence MLP Re-weighting
Peking University(北京大学);School of Science(理学院);Westlake University(西湖大学);Beijing International Center for Mathematical Research(北京国际数学研究中心);Center for Machine Learning Research(机器学习研究中心)
Medical Foundation Models are Susceptible to Targeted Misinformation Attacks
University Hospital Aachen(亚琛大学医院);University Medical Center of the Johannes Gutenberg-University(约翰内斯·古腾堡大学医学中心);University Medicine Essen(埃森大学医学部);Ocumeda GmbH(Ocumeda有限公司);Charité - Universitätsmedizin Berlin(柏林夏里特医学院);Freie Universität Berlin(柏林自由大学);Humboldt Universität zu Berlin(柏林洪堡大学);Berlin Institute of Health at Charité - Universitätsmedizin Berlin(柏林夏里特医学院柏林健康研究所);Technical University Dresden(德累斯顿工业大学);University Hospital Dresden(德累斯顿大学医院);University Hospital Heidelberg(海德堡大学医院)
StealthGraph: Exposing Domain-Specific Risks in LLMs through Knowledge-Graph-Guided Harmful Prompt Generation
School of Software Technology, Zhejiang University(浙江大学软件学院);College of Computer Science and Technology, Zhejiang University(浙江大学计算机科学与技术学院);State Key Lab of CAD&CG, Zhejiang University(浙江大学CAD&CG国家重点实验室)