An influential California legislator is pressuring the State Bar of California to ditch its new multiple-choice questions after a February bar exam debacle and revert to the traditional test format in July. "Given the catastrophe of the February bar, I think that going back to the methods that have been used for the last 50 years -- until we can adequately test what new methods may be employed -- is the appropriate way to go," Sen. Tom Umberg (D-Orange), chair of the state Senate Judiciary Committee, told The Times. Thousands of test takers seeking to practice law in California typically take the two-day bar exam in July. Reverting to the national system by the National Conference of Bar Examiners, which California has used since 1972, would be a major retreat for the embattled State Bar. Its new exam was rolled out this year as a cost-cutting measure and "historic agreement" that would offer test takers the choice of remote testing.

You can want different things from a university--superlative basketball, an arts center, competent instruction in philosophy or physics, even a cure for cancer. No wonder these institutions struggle to keep everyone happy. The Trump Administration has effectively declared open war on higher education, targeting it with deep cuts to federal grant funding. University presidents are alarmed, as are faculty members, and anyone who cares about the university's broader role. Because I'm a historian of science and technology, part of my terrain is the evolving role of the university--from its medieval, clerical origins to the entrepreneurial R. & D. engines of today.

We present a framework for adaptive-stepsize MCMC sampling based on time-rescaled Langevin dynamics, in which the stepsize variation is dynamically driven by an additional degree of freedom. Our approach augments the phase space by an additional variable which in turn defines a time reparameterization. The use of an auxiliary relaxation equation allows accumulation of a moving average of a local monitor function and provides for precise control of the timestep while circumventing the need to modify the drift term in the physical system. Our algorithm is straightforward to implement and can be readily combined with any off-the-peg fixed-stepsize Langevin integrator. As a particular example, we consider control of the stepsize by monitoring the norm of the log-posterior gradient, which takes inspiration from the Adam optimizer, the stepsize being automatically reduced in regions of steep change of the log posterior and increased on plateaus, improving numerical stability and convergence speed. As in Adam, the stepsize variation depends on the recent history of the gradient norm, which enhances stability and improves accuracy compared to more immediate control approaches. We demonstrate the potential benefit of this method--both in accuracy and in stability--in numerical experiments including Neal's funnel and a Bayesian neural network for classification of MNIST data.

The House of Representatives, Japan's lower chamber of parliament, passed a bill on Thursday to promote the development of artificial intelligence technology and take steps to mitigate its risks. The legislation is expected to be enacted during the current parliamentary session set to end in June after deliberations at the House of Councilors, the upper chamber. AI "will be the foundation of economic and social development and is an important technology from the viewpoint of security," the bill said.

Traditional fine-tuning methods, while effective, often require substantial computational resources and large, static datasets, making them impractical for real-time applications. Moreover, these models notoriously suffer from catastrophic forgetting, rapid performance degradation on previously learned tasks when presented with new data (Luo et al., 2023). Recent literature addresses catastrophic forgetting via techniques such as replay buffers, which periodically reintroduce previously learned data, and Low-Rank Adaptation (LoRA), a parameter-efficient fine-tuning approach designed to reduce computational overhead (Smith & Jones, 2024; Hu et al., 2021). Although these methods individually show promise, there remains a notable gap in understanding their efficacy and interaction within real-time, streaming learning environments. In this work, we bridge this gap by integrating LoRA with a lightweight replay mechanism under stringent streaming constraints, simulating real-world conditions where models must continually adapt using limited computational resources and data batches. We focus specifically on three distinct domains,medical, genetic, and legal,to evaluate the generalizability and robustness of our approach.

The proliferation of abusive language in online communications has posed significant risks to the health and wellbeing of individuals and communities. The growing concern regarding online abuse and its consequences necessitates methods for identifying and mitigating harmful content and facilitating continuous monitoring, moderation, and early intervention. This paper presents a taxonomy for distinguishing key characteristics of abusive language within online text. Our approach uses a systematic method for taxonomy development, integrating classification systems of 18 existing multi-label datasets to capture key characteristics relevant to online abusive language classification. The resulting taxonomy is hierarchical and faceted, comprising 5 categories and 17 dimensions. It classifies various facets of online abuse, including context, target, intensity, directness, and theme of abuse. This shared understanding can lead to more cohesive efforts, facilitate knowledge exchange, and accelerate progress in the field of online abuse detection and mitigation among researchers, policy makers, online platform owners, and other stakeholders.

Large language models (LLMs) often generate responses that deviate from user input or training data, a phenomenon known as "hallucination." These hallucinations undermine user trust and hinder the adoption of generative AI systems. Addressing hallucinations is essential for the advancement of LLMs. This paper introduces a comprehensive hallucination benchmark, incorporating both new extrinsic and existing intrinsic evaluation tasks, built upon clear taxonomy of hallucination. A major challenge in benchmarking hallucinations is the lack of a unified framework due to inconsistent definitions and categorizations. We disentangle LLM hallucination from "factuality," proposing a clear taxonomy that distinguishes between extrinsic and intrinsic hallucinations, to promote consistency and facilitate research. Extrinsic hallucinations, where the generated content is not consistent with the training data, are increasingly important as LLMs evolve. Our benchmark includes dynamic test set generation to mitigate data leakage and ensure robustness against such leakage. We also analyze existing benchmarks, highlighting their limitations and saturation. The work aims to: (1) establish a clear taxonomy of hallucinations, (2) introduce new extrinsic hallucination tasks, with data that can be dynamically regenerated to prevent saturation by leakage, (3) provide a comprehensive analysis of existing benchmarks, distinguishing them from factuality evaluations.

As generative AI models become increasingly integrated into high-stakes domains, the need for robust methods to evaluate their ethical reasoning becomes increasingly important. This paper introduces a five-dimensional audit model -- assessing Analytic Quality, Breadth of Ethical Considerations, Depth of Explanation, Consistency, and Decisiveness -- to evaluate the ethical logic of leading large language models (LLMs). Drawing on traditions from applied ethics and higher-order thinking, we present a multi-battery prompt approach, including novel ethical dilemmas, to probe the models' reasoning across diverse contexts. We benchmark seven major LLMs finding that while models generally converge on ethical decisions, they vary in explanatory rigor and moral prioritization. Chain-of-Thought prompting and reasoning-optimized models significantly enhance performance on our audit metrics. This study introduces a scalable methodology for ethical benchmarking of AI systems and highlights the potential for AI to complement human moral reasoning in complex decision-making contexts.

Abstract--T o jointly tackle the challenges of data and node heterogeneity in decentralized learning, we propose a dist ributed strong lottery ticket hypothesis (DSL TH), based on which a communication-efficient personalized learning algorithm is developed. In the proposed method, each local model is represente d as the Hadamard product of global real-valued parameters and a personalized binary mask for pruning. The local model is lea rned by updating and fusing the personalized binary masks while the real-valued parameters are fixed among different agents . T o further reduce the complexity of hardware implementatio n, we incorporate a group sparse regularization term in the los s function, enabling the learned local model to achieve struc - tured sparsity. Then, a binary mask aggregation algorithm i s designed by introducing an intermediate aggregation tenso r and adding a personalized fine-tuning step in each iteration, wh ich constrains model updates towards the local data distributi on. The proposed method effectively leverages the relativity a mong agents while meeting personalized requirements in heterog eneous node conditions. We also provide a theoretical proof for the DSL TH, establishing it as the foundation of the proposed met hod. Numerical simulations confirm the validity of the DSL TH and demonstrate the effectiveness of the proposed algorithm. Index T erms--Distributed learning, personalized learning, data and node heterogeneity, communication efficiency. As one of the most promising applications in 6G era, Artificial Intelligence of Things (AIoT) combines the artifi cial intelligence technologies with the Internet of Things (IoT) infrastructure, resembling the transformation from "conn ected things" to "connected intelligence" . This work was supported in part by National Natural Science F oundation of China under Grants 62394292 and U20A20158, Ministry of In dustry and Information Technology under Grant TC220H07E, Zhejiang Pr ovincial Key R&D Program under Grant 2023C01021, the Fundamental Resear ch Funds for the Central Universities No. 226-2024-00069, and the EU-SN S 6G CENTRIC Project. Z. Tian (email: dankotian@zju.edu.cn) was with the College of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China and now is with the Center for Wireless Communications, University of Oulu, Oulu 90014, Finland. Z. Zhang (Corresponding Author, email: ning ming@zju.edu.cn) is with the College of Information Science and Electronic Engineer ing, Zhejiang University, Hangzhou 310027, China, and with the State Key L aboratory of Industrial Control Technology, Hangzhou 310027, China, and also with Zhejiang Provincial Key Laboratory of Multimodal Communic ation Networks and Intelligent Information Processing, Hangzhou 310027, China.

Developing lifelong learning agents is crucial for artificial general intelligence. However, deep reinforcement learning (RL) systems often suffer from plasticity loss, where neural networks gradually lose their ability to adapt during training. Despite its significance, this field lacks unified benchmarks and evaluation protocols. We introduce Plasticine, the first open-source framework for benchmarking plasticity optimization in deep RL. Plasticine provides single-file implementations of over 13 mitigation methods, 10 evaluation metrics, and learning scenarios with increasing non-stationarity levels from standard to open-ended environments. This framework enables researchers to systematically quantify plasticity loss, evaluate mitigation strategies, and analyze plasticity dynamics across different contexts. Our documentation, examples, and source code are available at https://github.com/