The Tokyo District Public Prosecutor's Office's special investigation squad arrested Kazutaka Yonekura, 48, the founder and former president of Japanese artificial intelligence developer Alt, and three others on Thursday on suspicion of padding the firm's sales in violation of the financial instruments and exchange law. The other three include Yusuke Hioki, 34, also a former president of the Tokyo-based company. The special squad did not reveal whether the suspects have admitted the allegations against them. They allegedly submitted to the Kanto Local Finance Bureau in September 2024 financial statements, in which the company's sales in the period from January 2022 to June 2024 were inflated by about ¥8.4 billion. In March this year, after Alt's listing on the Tokyo Stock Exchange's Growth section for startup in October 2024, the suspects submitted a statement that overstated sales for the business year to December 2024 by about ¥4.9 billion, according to the special squad.

Japan's top bank CEOs push for AI, soothing worry over human work Japan's top financial leaders are working to ease fears that AI will cost jobs, emphasizing its role in boosting efficiency and transforming work. The heads of Japan's biggest financial firms are going out of their way to assuage worries that artificial intelligence will take away jobs. I don't think humans will lose their value. Humans have ability for dialogue, empathy, creativity and ethics," Mizuho Chief Executive Officer Masahiro Kihara said on Thursday at an event hosted by the Nikkei. People might say, 'what about my job if we use more AI?' I think they can aim for more value-added work."

Recent advances in Large Language Models (LLMs) and Reinforcement Learning (RL) have led to strong performance in open-domain question answering (QA). However, existing models still struggle with questions that admit multiple valid answers. Standard QA benchmarks, which typically assume a single gold answer, overlook this reality and thus produce inappropriate training signals. Existing attempts to handle ambiguity often rely on costly manual annotation, which is difficult to scale to multi-hop datasets such as HotpotQA and MuSiQue. In this paper, we present A$^2$Search, an annotation-free, end-to-end training framework to recognize and handle ambiguity. At its core is an automated pipeline that detects ambiguous questions and gathers alternative answers via trajectory sampling and evidence verification. The model is then optimized with RL using a carefully designed $\mathrm{AnsF1}$ reward, which naturally accommodates multiple answers. Experiments on eight open-domain QA benchmarks demonstrate that A$^2$Search achieves new state-of-the-art performance. With only a single rollout, A$^2$Search-7B yields an average $\mathrm{AnsF1}@1$ score of $48.4\%$ across four multi-hop benchmarks, outperforming all strong baselines, including the substantially larger ReSearch-32B ($46.2\%$). Extensive analyses further show that A$^2$Search resolves ambiguity and generalizes across benchmarks, highlighting that embracing ambiguity is essential for building more reliable QA systems. Our code, data, and model weights can be found at https://github.com/zfj1998/A2Search

Knowledge conflict often arises in retrieval-augmented generation (RAG) systems, where retrieved documents may be inconsistent with one another or contradict the model's parametric knowledge. Existing benchmarks for investigating the phenomenon have notable limitations, including a narrow focus on the question answering setup, heavy reliance on entity substitution techniques, and a restricted range of conflict types. To address these issues, we propose a knowledge graph (KG)-based framework that generates varied and subtle conflicts between two similar yet distinct contexts, while ensuring interpretability through the explicit relational structure of KGs. Experimental results on our benchmark, MAGIC, provide intriguing insights into the inner workings of LLMs regarding knowledge conflict: both open-source and proprietary models struggle with conflict detection -- especially when multi-hop reasoning is required -- and often fail to pinpoint the exact source of contradictions. Finally, we present in-depth analyses that serve as a foundation for improving LLMs in integrating diverse, sometimes even conflicting, information.

In their paper Policy Design for Two-sided Platforms with Participation Dynamics, which was presented at ICML 2025, and investigated the the participation dynamics in two-sided markets. In this interview, Haruka tells us more about such two-sided platforms, the main contributions of the work, and the experiments carried out to test the method. What is the topic of the research in your paper and why is it an interesting area for study? Our paper studied the long-term impacts of decision-making algorithms on two-sided platforms like e-commerce or music streaming applications. In two-sided platforms, multiple stakeholders, such as viewers and content creators, are involved.

API calls for each position identified in a piece of text. Question Answering We use the Atlas model of Izacard et al. (2022) finetuned on Natural Questions Calculator Our calculator is based on a simple Python script and only supports the operators " It does not return any result for syntactically invalid equations. "=", "equals", "equal to", "total of", "average of" followed by a number, or (iii) contain at least three English text before generating API calls. Below, we list the prompts used to sample API calls for each tool considered. Your task is to add calls to a Question Answering API to a piece of text. Input: Joe Biden was born in Scranton, Pennsylvania. Output: Joe Biden was born in [QA("Where was Joe Biden born?")] Scranton, [QA("In Output: Coca-Cola, or [QA("What other name is Coca-Cola known by?")] Coke, is Your task is to add calls to a Calculator API to a piece of text.

A China Coast Guard vessel is seen on a giant screen showing news footage about the coast guard's law enforcement patrols in waters around Taiwan, outside a shopping mall in Beijing on April 1. | REUTERS TAIPEI - China is increasing military activities near Taiwan and honing its ability to stage a surprise attack, as well as seeking to undermine trust in the government with hybrid online warfare tactics, the island's defense ministry said on Thursday. Democratically-governed Taiwan, which China views as its own territory, has faced increased military pressure from Beijing over the past five years, including at least seven rounds of major war games around the island since 2022. China has been using artificial intelligence tools to weaken Taiwan's cybersecurity and to scan for weak points in critical infrastructure, the defense ministry said in a report released every two years. Beijing is also using hybrid warfare to weaken people's trust in the government and support for defense spending, and stepping up grey zone harassment, it added, referring to non-combat operations such as coast guard patrols designed to pressure Taiwan. Through both conventional and unconventional military actions, it aims to test its capabilities for attacking Taiwan and confronting foreign forces, the ministry said.

Large language models (LLMs) have been recently leveraged as training data generators for various natural language processing (NLP) tasks. While previous research has explored different approaches to training models using generated data, they generally rely on simple class-conditional prompts, which may limit the diversity of the generated data and inherit systematic biases of LLM. Thus, we investigate training data generation with diversely attributed prompts (e.g.,