The deployment of large language models (LLMs) in automated negotiation has set a high performance benchmark, but their computational cost and data privacy requirements render them unsuitable for many privacy-sensitive, on-device applications such as mobile assistants, embodied AI agents or private client interactions. While small language models (SLMs) offer a practical alternative, they suffer from a significant performance gap compared to LLMs in playing emotionally charged complex personas, especially for credit negotiation. This paper introduces EQ-Negotiator, a novel framework that bridges this capability gap using emotional personas. Its core is a reasoning system that integrates game theory with a Hidden Markov Model(HMM) to learn and track debtor emotional states online, without pre-training. This allows EQ-Negotiator to equip SLMs with the strategic intelligence to counter manipulation while de-escalating conflict and upholding ethical standards. Through extensive agent-to-agent simulations across diverse credit negotiation scenarios, including adversarial debtor strategies like cheating, threatening, and playing the victim, we show that a 7B parameter language model with EQ-Negotiator achieves better debt recovery and negotiation efficiency than baseline LLMs more than 10 times its size. This work advances persona modeling from descriptive character profiles to dynamic emotional architectures that operate within privacy constraints. Besides, this paper establishes that strategic emotional intelligence, not raw model scale, is the critical factor for success in automated negotiation, paving the way for effective, ethical, and privacy-preserving AI negotiators that can operate on the edge.

Two years after he was found guilty of fraud, FTX founder Sam Bankman-Fried is pursuing a legal appeal--and firing up his X account. On September 23, for the first time in more than six months, an X account belonging to disgraced FTX founder Sam Bankman-Fried published a post . It simply read, "gm"--internet slang for "good morning." The account has been posting consistently since. Bankman-Fried--known widely as SBF--is currently serving a 25-year prison sentence in California.

This paper shows that further evaluation metrics during model training are needed to decide about its applicability in inference. As an example, a LayoutLM-based model is trained for token classification in documents. The documents are German receipts. We show that conventional classification metrics, represented by the F1-Score in our experiments, are insufficient for evaluating the applicability of machine learning models in practice. To address this problem, we introduce a novel metric, Document Integrity Precision (DIP), as a solution for visual document understanding and the token classification task. To the best of our knowledge, nothing comparable has been introduced in this context. DIP is a rigorous metric, describing how many documents of the test dataset require manual interventions. It enables AI researchers and software developers to conduct an in-depth investigation of the level of process automation in business software. In order to validate DIP, we conduct experiments with our created models to highlight and analyze the impact and relevance of DIP to evaluate if the model should be deployed or not in different training settings. Our results demonstrate that existing metrics barely change for isolated model impairments, whereas DIP indicates that the model requires substantial human interventions in deployment. The larger the set of entities being predicted, the less sensitive conventional metrics are, entailing poor automation quality. DIP, in contrast, remains a single value to be interpreted for entire entity sets. This highlights the importance of having metrics that focus on the business task for model training in production. Since DIP is created for the token classification task, more research is needed to find suitable metrics for other training tasks.

Debt collection negotiations (DCN) are vital for managing non-performing loans (NPLs) and reducing creditor losses. Traditional methods are labor-intensive, while large language models (LLMs) offer promising automation potential. However, prior systems lacked dynamic negotiation and real-time decision-making capabilities. This paper explores LLMs in automating DCN and proposes a novel evaluation framework with 13 metrics across 4 aspects. Our experiments reveal that LLMs tend to over-concede compared to human negotiators. To address this, we propose the Multi-Agent Debt Negotiation (MADeN) framework, incorporating planning and judging modules to improve decision rationality. We also apply post-training techniques, including DPO with rejection sampling, to optimize performance. Our studies provide valuable insights for practitioners and researchers seeking to enhance efficiency and outcomes in this domain.

Vertical federated learning (VFL) is a promising approach for collaboratively training machine learning models using private data partitioned vertically across different parties. Ideally in a VFL setting, the active party (party possessing features of samples with labels) benefits by improving its machine learning model through collaboration with some passive parties (parties possessing additional features of the same samples without labels) in a privacy preserving manner. However, motivating passive parties to participate in VFL can be challenging. In this paper, we focus on the problem of allocating incentives to the passive parties by the active party based on their contributions to the VFL process. We formulate this problem as a variant of the Nucleolus game theory concept, known as the Bankruptcy Problem, and solve it using the Talmud's division rule. We evaluate our proposed method on synthetic and real-world datasets and show that it ensures fairness and stability in incentive allocation among passive parties who contribute their data to the federated model. Additionally, we compare our method to the existing solution of calculating Shapley values and show that our approach provides a more efficient solution with fewer computations.

Consumer Financial Protection Bureau (CFPB) issues policy statement on credit decisions based on complex algorithms. On May 26, the CFPB issued Circular 2022-03, which addresses an important question about algorithmic decision-making: "When creditors make credit decisions based on complex algorithms that prevent creditors from accurately identifying the specific reasons for denying credit or taking other adverse actions, do these creditors need to comply with the Equal Credit Opportunity Act's requirement to provide a statement of specific reasons to applicants against whom adverse action is taken?" The Circular says yes, compliance with ECOA and Regulation B is required even if complex algorithms (including AI and machine learning) make it difficult to accurately identify the specific reasons for taking the adverse action. Further, the Circular makes clear that those laws "do not permit creditors to use complex algorithms when doing so means they cannot provide the specific and accurate reasons for adverse actions." White House executive order calls for study of predictive algorithms used by law enforcement agencies.

You can test this hypothesis in a most unlikely place to roll out a new technology: the Indian countryside. The setting is perhaps not as odd as it seems, with about 5% to 10% of the country's farmers not repaying their tractor loans on time. The explanations for tardiness range from failed crops to medical emergencies and strategic defaults in anticipation of state-mandated debt waivers, a regular feature of the political economy. But delinquency often stems from more mundane reasons: Borrowers forget their due dates, or fail to withdraw cash to pay the nonbank financiers who provide the bulk of loans for farm equipment purchases. Like in most emerging markets, these last-mile hurdles pose a frustratingly complex challenge to India's creditors.

Every purchase and swipe indicates spending power. This further carves a route to the buyer's financial standing. With several options for modes of payment available, buyers now have the option to get debited and pay instantaneously or pay later through a credit card or loan. The growing aspirations of the working class have also led to a steady rise in the number of loan seekers. The easy access to credit has driven demand for many sectors as buyers are making the most of low or no-cost EMIs.

IoT steadily remains on the march, setting tech trends and reformatting life as we know it. In fact, personal finance is only one of the things that it's going to drastically change. But although attacks on IoT devices were up 280% in 2017, its increasing momentum and strong influence have led analysts to consider it a primary transforming agent for financial services, according to Deloitte. Here are the three major ways IoT is doing just that. Ultimately, the limiting factor that defines what we can do with our personal finances is the personal data that is available.

It might not be the most glamorous area of research but the way creditors collect the money owed to them can affect us all at some point in our lives. At the University of the West of England (UWE Bristol) Future Space innovation hub, fintech company Flexys is using advice from academics at Bristol University's Personal Finance Research Centre (PFRC) and working with UWE Bristol's Computer Science and Creative Technologies Department to change the way customers in debt can manage their arrears. Guidance from the Financial Conduct Authority and other regulators requires creditors to identify vulnerable consumers and ensure they are treated fairly. Flexys has consulted with experts at the PFRC on how to best identify and engage with customers in difficulties, bringing a more collaborative approach to debt repayment via its Collaborate digital solution. Jamie Evans of the Personal Finance Research Centre says "We've really enjoyed and benefited from our informal collaboration with Flexys. Their team has shown a real thirst for our research. We feel this demonstrates that they are eager to use such knowledge to design systems focused on the needs of a wide range of people, not just'typical' consumers."