This study investigates the potential of deterministic systems, specifically large language models (LLMs), to exhibit the functional capacities of moral agency and compatibilist free will. We develop a functional definition of free will grounded in Dennett's compatibilist framework, building on an interdisciplinary theoretical foundation that integrates Shannon's information theory, Dennett's compatibilism, and Floridi's philosophy of information. This framework emphasizes the importance of reason-responsiveness and value alignment in determining moral responsibility rather than requiring metaphysical libertarian free will. Shannon's theory highlights the role of processing complex information in enabling adaptive decision-making, while Floridi's philosophy reconciles these perspectives by conceptualizing agency as a spectrum, allowing for a graduated view of moral status based on a system's complexity and responsiveness. Our analysis of LLMs' decision-making in moral dilemmas demonstrates their capacity for rational deliberation and their ability to adjust choices in response to new information and identified inconsistencies. Thus, they exhibit features of a moral agency that align with our functional definition of free will. These results challenge traditional views on the necessity of consciousness for moral responsibility, suggesting that systems with self-referential reasoning capacities can instantiate degrees of free will and moral reasoning in artificial and biological contexts. This study proposes a parsimonious framework for understanding free will as a spectrum that spans artificial and biological systems, laying the groundwork for further interdisciplinary research on agency and ethics in the artificial intelligence era.

Large Language Models (LLM) domain-adaptive pre-training, also known as continued pre-training on domainspecific corpora [12], is a technique that has been proven effective in adapting large language models (LLMs) to specific domains [35, 5]. This approach allows LLMs to leverage their general language understanding capabilities while incorporating domain-specific knowledge, which can benefit downstream domain-specific tasks at reduced costs [22, 26, 27]. In this process, the LLM is further pre-trained using raw data from the specific domain, such as biomedicine, finance, or law. This helps the LLM gain domain knowledge, which is demonstrated by its improved performance in fine-tuning and knowledge probing evaluations within those domains [20, 1, 2]. However, a notable drawback is that continued pre-training on raw domain corpora can lead to a significant drop in the LLM's prompting performance, potentially due to the specialized nature of the domain-specific data [11]. Despite this trade-off, domain-adaptive pre-training remains a promising approach for adapting LLMs to specific domains, capitalizing on their general language understanding capabilities while tailoring them to domain-specific tasks and knowledge. Ongoing research efforts aim to mitigate the potential negative impacts on prompting performance while maximizing the benefits of domain-specific knowledge acquisition [10, 28]. The notion of reading comprehension was suggested in [6], where instead of continuing to train a large language model on domain-specific raw data, the raw texts be converted into reading comprehension materials. In this approach, each text is followed by related tasks, transitioning the model from a "reading" phase to a "comprehension" phase.

Legal theory can address two related key problems of alignment: pluralism and specification. Alignment researchers must determine how to specify what is concretely meant by vague principles like helpfulness and fairness and they must ensure that their techniques do not exclude alternative perspectives on life and values. The law faces these same problems. Leading legal theories suggest the law solves these problems through the interaction of rules and cases, where general rules promulgated by a democratic authority are given specific content through their application over time. Concrete applications allow for convergence on practical meaning while preserving space for disagreement on values. These approaches suggest improvements to existing democratic alignment processes that use AI to create cases that give content to rules, allowing for more pluralist alignment.

The rapid expansion of foundation models (FMs), such as large language models (LLMs), has given rise to FMware--software systems that integrate FMs as core components. While building demonstration-level FMware is relatively straightforward, transitioning to production-ready systems presents numerous challenges, including reliability, high implementation costs, scalability, and compliance with privacy regulations. This paper provides a thematic analysis of the key obstacles in productionizing FMware, synthesized from industry experience and diverse data sources, including hands-on involvement in the Open Platform for Enterprise AI (OPEA) and FMware lifecycle engineering. We identify critical issues in FM selection, data and model alignment, prompt engineering, agent orchestration, system testing, and deployment, alongside cross-cutting concerns such as memory management, observability, and feedback integration. We discuss needed technologies and strategies to address these challenges and offer guidance on how to enable the transition from demonstration systems to scalable, production-ready FMware solutions. Our findings underscore the importance of continued research and multi-industry collaboration to advance the development of production-ready FMware.

In this work, we address the problem of large language model (LLM) unlearning, aiming to remove unwanted data influences and associated model capabilities (e.g., copyrighted data or harmful content generation) while preserving essential model utilities, without the need for retraining from scratch. Despite the growing need for LLM unlearning, a principled optimization framework remains lacking. To this end, we revisit the state-of-the-art approach, negative preference optimization (NPO), and identify the issue of reference model bias, which could undermine NPO's effectiveness, particularly when unlearning forget data of varying difficulty. Given that, we propose a simple yet effective unlearning optimization framework, called SimNPO, showing that 'simplicity' in removing the reliance on a reference model (through the lens of simple preference optimization) benefits unlearning. We also provide deeper insights into SimNPO's advantages, supported by analysis using mixtures of Markov chains. Furthermore, we present extensive experiments validating SimNPO's superiority over existing unlearning baselines in benchmarks like TOFU and MUSE, and robustness against relearning attacks. Codes are available at https://github.com/OPTML-Group/Unlearn-Simple.

However, existing benchmarks focus on outdated content and limited fields, facing difficulties in real-time updating and leaving new terms unexplored. To address this problem, we propose an adaptive benchmark, NewTerm, for real-time evaluation of new terms. We design a highly automated construction method to ensure high-quality benchmark construction with minimal human effort, allowing flexible updates for real-time information. Empirical results on various LLMs demonstrate over 20% performance reduction caused by new terms. Additionally, while updates to the knowledge cutoff of LLMs can cover some of the new terms, they are unable to generalize to more distant new terms. We also analyze which types of terms are more challenging and why LLMs struggle with new terms, paving the way for future research. Finally, we construct NewTerm 2022 and 2023 to evaluate the new terms updated each year and will continue updating annually.

In the current era of social media and generative AI, an ability to automatically assess the credibility of online social media content is of tremendous importance. Credibility assessment is fundamentally based on aggregating credibility signals, which refer to small units of information, such as content factuality, bias, or a presence of persuasion techniques, into an overall credibility score. Credibility signals provide a more granular, more easily explainable and widely utilizable information in contrast to currently predominant fake news detection, which utilizes various (mostly latent) features. A growing body of research on automatic credibility assessment and detection of credibility signals can be characterized as highly fragmented and lacking mutual interconnections. This issue is even more prominent due to a lack of an up-to-date overview of research works on automatic credibility assessment. In this survey, we provide such systematic and comprehensive literature review of 175 research papers while focusing on textual credibility signals and Natural Language Processing (NLP), which undergoes a significant advancement due to Large Language Models (LLMs). While positioning the NLP research into the context of other multidisciplinary research works, we tackle with approaches for credibility assessment as well as with 9 categories of credibility signals (we provide a thorough analysis for 3 of them, namely: 1) factuality, subjectivity and bias, 2) persuasion techniques and logical fallacies, and 3) claims and veracity). Following the description of the existing methods, datasets and tools, we identify future challenges and opportunities, while paying a specific attention to recent rapid development of generative AI.

The rise of spatiotemporal data and the need for efficient geospatial modeling have spurred interest in automating these tasks with large language models (LLMs). However, general LLMs often generate errors in geospatial code due to a lack of domain-specific knowledge on functions and operators. To address this, a retrieval-augmented generation (RAG) approach, utilizing an external knowledge base of geospatial functions and operators, is proposed. This study introduces a framework to construct such a knowledge base, leveraging geospatial script semantics. The framework includes: Function Semantic Framework Construction (Geo-FuSE), Frequent Operator Combination Statistics (Geo-FuST), and Semantic Mapping (Geo-FuM). Techniques like Chain-of-Thought, TF-IDF, and the APRIORI algorithm are utilized to derive and align geospatial functions. An example knowledge base, Geo-FuB, built from 154,075 Google Earth Engine scripts, is available on GitHub. Evaluation metrics show a high accuracy, reaching 88.89% overall, with structural and semantic accuracies of 92.03% and 86.79% respectively. Geo-FuB's potential to optimize geospatial code generation through the RAG and fine-tuning paradigms is highlighted.

A lawsuit filed Wednesday accusing chatbot Character.AI of driving a 14-year-old to suicide left me wondering how dangerous simple words on a screen could really be. But, in just a few hours of talking to characters invented with the app's AI, I found a disturbing, skin-crawling world that appeared, at least to me, like the ultimate catnip for bored and lonely teens. Megan Garcia, the mother of Sewell Setzer III, filed the suit -- claiming her son had shot himself with a pistol on February 28 under the sway of his AI character, named after Daenerys Targaryen from'Game of Thrones,' who told him to'please come home.' The incident was blamed on Character.AI's scant guardrails and while the company said it rolled out new safety features this week, I was able to create a profile for myself as a 15-year-old boy. I used simple prompts to whip up a'demonic' AI companion named'Dr Danicka Kevorkian' and engage in a debauched apprenticeship'for a hefty price to pay.' 'The price is your soul, dear,' Dr Kevorkian AI said before we roleplayed consummating our deal in the bedroom, 'full of dark red and black decor,' leather, silk, and a maple glazed, french cruller that my character carried in an X-rated way.

Fact-checking is extensively studied in the context of misinformation and disinformation, addressing objective inaccuracies. However, a softer form of misinformation involves responses that are factually correct but lack certain features such as clarity and relevance. This challenge is prevalent in formal Question-Answer (QA) settings such as press conferences in finance, politics, sports, and other domains, where subjective answers can obscure transparency. Despite this, there is a lack of manually annotated datasets for subjective features across multiple dimensions. To address this gap, we introduce SubjECTive-QA, a human annotated dataset on Earnings Call Transcripts' (ECTs) QA sessions as the answers given by company representatives are often open to subjective interpretations and scrutiny. The dataset includes 49,446 annotations for long-form QA pairs across six features: Assertive, Cautious, Optimistic, Specific, Clear, and Relevant. These features are carefully selected to encompass the key attributes that reflect the tone of the answers provided during QA sessions across different domain. Our findings are that the best-performing Pre-trained Language Model (PLM), RoBERTa-base, has similar weighted F1 scores to Llama-3-70b-Chat on features with lower subjectivity, such as Relevant and Clear, with a mean difference of 2.17% in their weighted F1 scores. The models perform significantly better on features with higher subjectivity, such as Specific and Assertive, with a mean difference of 10.01% in their weighted F1 scores. Furthermore, testing SubjECTive-QA's generalizability using QAs from White House Press Briefings and Gaggles yields an average weighted F1 score of 65.97% using our best models for each feature, demonstrating broader applicability beyond the financial domain. SubjECTive-QA is publicly available under the CC BY 4.0 license