Large language models (LLMs) like ChatGPT, Gemini, or LLaMA have been trending recently, demonstrating considerable advancement and generalizability power in countless domains. However, LLMs create an even bigger black box exacerbating opacity, with interpretability limited to few approaches. The uncertainty and opacity embedded in LLMs' nature restrict their application in high-stakes domains like financial fraud, phishing, etc. Current approaches mainly rely on traditional textual classification with posterior interpretable algorithms, suffering from attackers who may create versatile adversarial samples to break the system's defense, forcing users to make trade-offs between efficiency and robustness. To address this issue, we propose a novel cascading framework called Genshin (General Shield for Natural Language Processing with Large Language Models), utilizing LLMs as defensive one-time plug-ins. Unlike most applications of LLMs that try to transform text into something new or structural, Genshin uses LLMs to recover text to its original state. Genshin aims to combine the generalizability of the LLM, the discrimination of the median model, and the interpretability of the simple model. Our experiments on the task of sentimental analysis and spam detection have shown fatal flaws of the current median models and exhilarating results on LLMs' recovery ability, demonstrating that Genshin is both effective and efficient. In our ablation study, we unearth several intriguing observations. Utilizing the LLM defender, a tool derived from the 4th paradigm, we have reproduced BERT's 15% optimal mask rate results in the 3rd paradigm of NLP. Additionally, when employing the LLM as a potential adversarial tool, attackers are capable of executing effective attacks that are nearly semantically lossless.

Large Language Models (LLMs) have shown their impressive capabilities, while also raising concerns about the data contamination problems due to privacy issues and leakage of benchmark datasets in the pre-training phase. Therefore, it is vital to detect the contamination by checking whether an LLM has been pre-trained on the target texts. Recent studies focus on the generated texts and compute perplexities, which are superficial features and not reliable. In this study, we propose to utilize the probing technique for pre-training data detection by examining the model's internal activations. Our method is simple and effective and leads to more trustworthy pre-training data detection. Additionally, we propose ArxivMIA, a new challenging benchmark comprising arxiv abstracts from Computer Science and Mathematics categories. Our experiments demonstrate that our method outperforms all baselines, and achieves state-of-the-art performance on both WikiMIA and ArxivMIA, with additional experiments confirming its efficacy (Our code and dataset are available at https://github.com/zhliu0106/probing-lm-data).

To some, the advent of artificial intelligence (AI) promises better decision-making and increased military effectiveness while reducing the influence of human error and emotions. However, there is still debate about how AI systems, especially large language models (LLMs), behave compared to humans in high-stakes military decision-making scenarios with the potential for increased risks towards escalation and unnecessary conflicts. To test this potential and scrutinize the use of LLMs for such purposes, we use a new wargame experiment with 107 national security experts designed to look at crisis escalation in a fictional US-China scenario and compare human players to LLM-simulated responses in separate simulations. Wargames have a long history in the development of military strategy and the response of nations to threats or attacks. Here, we show a considerable high-level agreement in the LLM and human responses and significant quantitative and qualitative differences in individual actions and strategic tendencies. These differences depend on intrinsic biases in LLMs regarding the appropriate level of violence following strategic instructions, the choice of LLM, and whether the LLMs are tasked to decide for a team of players directly or first to simulate dialog between players. When simulating the dialog, the discussions lack quality and maintain a farcical harmony. The LLM simulations cannot account for human player characteristics, showing no significant difference even for extreme traits, such as "pacifist" or "aggressive sociopath". Our results motivate policymakers to be cautious before granting autonomy or following AI-based strategy recommendations.

Arman Rahmim Departments of Radiology and Physics, University of British Columbia Tyler J. Bradshaw Department of Radiology, University of Wisconsin Guido Davidzon Department of Radiology, Division of Nuclear Medicine & Molecular Imaging, Stanford University Joyita Dutta Department of Biomedical Engineering, University of Massachusetts Amherst Georges El Fakhri PET Center, Departments of Radiology & Biomedical Engineering and Bioinformatics & Data Science, Yale University Munir Ghesani United Theranostics Nicolas A. Karakatsanis Department of Radiology, Weill Cornell Medical College of Cornell University, New York Quanzheng Li Center for Advanced Medical Computing and Analysis, Department of Radiology, Massachusetts General Hospital, Harvard Medical School Chi Liu Department of Radiology and Biomedical Imaging, Yale University Emilie Roncali Departments of Biomedical Engineering and Radiology, University of California, Davis Babak Saboury Department of Radiology and Imaging Sciences, Clinical Center, National Institutes of Health Tahir Yusufaly Russell H. Morgan Department of Radiology and Radiological Sciences, Johns Hopkins School of Medicine Abhinav K. Jha Department of Biomedical Engineering and Mallinckrodt Institute of Radiology, Washington University, St. Louis Abstract The 2nd SNMMI Artificial Intelligence (AI) Summit, organized by the SNMMI AI Task Force, took place in Bethesda, MD, on February 29 - March 1, 2024. Bringing together various community members and stakeholders, and following up on a prior successful 2022 AI Summit, the summit theme was "AI in Action". Six key topics included (i) an overview of prior and ongoing efforts by the AI task force, (ii) emerging needs and tools for computational nuclear oncology, (iii) new frontiers in large language and generative models, (iv) defining the value proposition for the use of AI in nuclear medicine, (v) open science including efforts for data and model repositories, and (vi) issues of reimbursement and funding. The primary efforts, findings, challenges, and next steps are summarized in this manuscript. Introduction The Society of Nuclear Medicine & Molecular Imaging (SNMMI) 2nd Artificial Intelligence (AI) Summit, organized by the SNMMI AI Task Force, took place in Bethesda, MD, on February 29 - March 1, 2024. Over 100 community members and stakeholders from academia, healthcare, industry, and NIH gathered to discuss the emerging role of AI in nuclear medicine. It featured two plenaries, panel discussions, talks from leading experts in the field, and was concluded by a round table discussion on key findings, next steps, and call to action.

This study explores the risk preferences of Large Language Models (LLMs) and how the process of aligning them with human ethical standards influences their economic decision-making. By analyzing 30 LLMs, we uncover a broad range of inherent risk profiles ranging from risk-averse to risk-seeking. We then explore how different types of AI alignment, a process that ensures models act according to human values and that focuses on harmlessness, helpfulness, and honesty, alter these base risk preferences. Alignment significantly shifts LLMs towards risk aversion, with models that incorporate all three ethical dimensions exhibiting the most conservative investment behavior. Replicating a prior study that used LLMs to predict corporate investments from company earnings call transcripts, we demonstrate that although some alignment can improve the accuracy of investment forecasts, excessive alignment results in overly cautious predictions. These findings suggest that deploying excessively aligned LLMs in financial decision-making could lead to severe underinvestment. We underline the need for a nuanced approach that carefully balances the degree of ethical alignment with the specific requirements of economic domains when leveraging LLMs within finance.

Automatic evaluation methods for large language models (LLMs) are hindered by data contamination, leading to inflated assessments of their effectiveness. Existing strategies, which aim to detect contaminated texts, focus on quantifying contamination status instead of accurately gauging model performance. In this paper, we introduce KIEval, a Knowledge-grounded Interactive Evaluation framework, which incorporates an LLM-powered "interactor" role for the first time to accomplish a dynamic contamination-resilient evaluation. Starting with a question in a conventional LLM benchmark involving domain-specific knowledge, KIEval utilizes dynamically generated, multi-round, and knowledge-focused dialogues to determine whether a model's response is merely a recall of benchmark answers or demonstrates a deep comprehension to apply knowledge in more complex conversations. Extensive experiments on seven leading LLMs across five datasets validate KIEval's effectiveness and generalization. We also reveal that data contamination brings no contribution or even negative effect to models' real-world applicability and understanding, and existing contamination detection methods for LLMs can only identify contamination in pre-training but not during supervised fine-tuning.

Human preference alignment is essential to improve the interaction quality of large language models (LLMs). Existing alignment methods depend on manually annotated preference data to guide the LLM optimization directions. However, continuously updating LLMs for alignment raises a distribution gap between model-generated samples and human-annotated responses, hindering training effectiveness. To mitigate this issue, previous methods require additional preference annotation on newly generated samples to adapt to the shifted distribution, which consumes a large amount of annotation resources. Targeting more efficient human preference optimization, we propose an Adversarial Preference Optimization (APO) framework, in which the LLM and the reward model update alternatively via a min-max game. Through adversarial training, the reward model can adapt to the shifted generation distribution of the LLM without any additional annotation. With comprehensive experiments, we find the proposed adversarial training framework further enhances existing alignment baselines in terms of LLM helpfulness and harmlessness. The code is at https://github.com/Linear95/APO.

Large Language Models (LLMs) have demonstrated remarkable success in tasks like the Winograd Schema Challenge (WSC), showcasing advanced textual common-sense reasoning. However, applying this reasoning to multimodal domains, where understanding text and images together is essential, remains a substantial challenge. To address this, we introduce WinoVis, a novel dataset specifically designed to probe text-to-image models on pronoun disambiguation within multimodal contexts. Utilizing GPT-4 for prompt generation and Diffusion Attentive Attribution Maps (DAAM) for heatmap analysis, we propose a novel evaluation framework that isolates the models' ability in pronoun disambiguation from other visual processing challenges. Evaluation of successive model versions reveals that, despite incremental advancements, Stable Diffusion 2.0 achieves a precision of 56.7% on WinoVis, only marginally surpassing random guessing. Further error analysis identifies important areas for future research aimed at advancing text-to-image models in their ability to interpret and interact with the complex visual world.

The advancement of large language models (LLMs) has propelled the development of dialogue systems. Unlike the popular ChatGPT-like assistant model, which only satisfies the user's preferences, task-oriented dialogue systems have also faced new requirements and challenges in the broader business field. They are expected to provide correct responses at each dialogue turn, at the same time, achieve the overall goal defined by the task. By understanding rhetorical structures and topic structures via topic segmentation and discourse parsing, a dialogue system may do a better planning to achieve both objectives. However, while both structures belong to discourse structure in linguistics, rhetorical structure and topic structure are mostly modeled separately or with one assisting the other in the prior work. The interaction between these two structures has not been considered for joint modeling and mutual learning. Furthermore, unsupervised learning techniques to achieve the above are not well explored. To fill this gap, we propose an unsupervised mutual learning framework of two structures leveraging the global and local connections between them. We extend the topic modeling between non-adjacent discourse units to ensure global structural relevance with rhetorical structures. We also incorporate rhetorical structures into the topic structure through a graph neural network model to ensure local coherence consistency. Finally, we utilize the similarity between the two fused structures for mutual learning. The experimental results demonstrate that our methods outperform all strong baselines on two dialogue rhetorical datasets (STAC and Molweni), as well as dialogue topic datasets (Doc2Dial and TIAGE). We provide our code at https://github.com/Jeff-Sue/URT.

Privacy is a human right. It ensures that individuals are free to engage in discussions, participate in groups, and form relationships online or offline without fear of their data being inappropriately harvested, analyzed, or otherwise used to harm them. Preserving privacy has emerged as a critical factor in research, particularly in the computational social science (CSS), artificial intelligence (AI) and data science domains, given their reliance on individuals' data for novel insights. The increasing use of advanced computational models stands to exacerbate privacy concerns because, if inappropriately used, they can quickly infringe privacy rights and lead to adverse effects for individuals -- especially vulnerable groups -- and society. We have already witnessed a host of privacy issues emerge with the advent of large language models (LLMs), such as ChatGPT, which further demonstrate the importance of embedding privacy from the start. This article contributes to the field by discussing the role of privacy and the issues that researchers working in CSS, AI, data science and related domains are likely to face. It then presents several key considerations for researchers to ensure participant privacy is best preserved in their research design, data collection and use, analysis, and dissemination of research results.