Large Reasoning Models (LRMs) exhibit remarkable reasoning abilities but rely primarily on parametric knowledge, limiting factual accuracy. While recent works equip reinforcement learning (RL)-based LRMs with retrieval capabilities, they suffer from overthinking and lack robustness in reasoning, reducing their effectiveness in question answering (QA) tasks. To address this, we propose ReaRAG, a factuality-enhanced reasoning model that explores diverse queries without excessive iterations. Our solution includes a novel data construction framework with an upper bound on the reasoning chain length. Specifically, we first leverage an LRM to generate deliberate thinking, then select an action from a predefined action space (Search and Finish). For Search action, a query is executed against the RAG engine, where the result is returned as observation to guide reasoning steps later. This process iterates until a Finish action is chosen. Benefiting from ReaRAG's strong reasoning capabilities, our approach outperforms existing baselines on multi-hop QA. Further analysis highlights its strong reflective ability to recognize errors and refine its reasoning trajectory. Our study enhances LRMs' factuality while effectively integrating robust reasoning for Retrieval-Augmented Generation (RAG).

Artificial intelligence is raising the possibility that students could cheat when writing papers. But educators and technology companies say they are ahead of the curve. Since its launch in November, the artificial intelligence-based program ChatGPT has drawn a lot of attention for its ability to quickly generate written passages based on simple prompts. Tell it to write you a 500-word essay on "The Old Man and The Sea," and within moments, you have a completed assignment that may have taken a student hours to write. With so much attention has come a lot of criticism and concern, especially in the realm of education.

For many people, the phrase "artificial intelligence" conjures up images of human-like robots and self-driving vehicles. For Kerrie Holley, AI has a much more human meaning. Holley had a long career, first as a distinguished engineer and later IBM fellow, before joining Eden Prairie-based Optum, a UnitedHealth Group unit that provides information and technology-based health services. At Optum, he's focused on developing technologies that make the health system work better and people's lives healthier. AI is one of those technologies.

Kerrie Holley has had a storied career in the technology industry spending nearly three decades at IBM before moving to Cisco as a chief technology officer focused on advancing the company's analytics and automation software platform In 2016, he became the first Technical Fellow for UnitedHealthcare Group's Optum division, responsible for steering strategy in emerging technologies like AI and machine learning, genomics and blockchain. MedCity News met with Holley during a recent trip to San Francico to discuss the potential of big data in healthcare, some of the barriers to adoption what can be done to win back the trust of patients. This exchange has been lightly edited for content and clarity. MedCity: What was the transition like moving from the traditional tech industry into healthcare? Holley: It's interesting because I'm working at a level that sees the shortfalls, sees the gaps and believes very strongly that we can make a difference with technology.

Optum, the $91 billion business within UnitedHealth Group, has its hands on a lot of information, from clinical data to information about healthcare consumers. The organization on its own has 140,000 employees, who work with 124 million members and 300 health plans. Now, it's exploring what it can do with that all information to make people healthier using artificial intelligence, an endeavor titled OptumIQ. "It's the coming together of the data, the analytics, and the expertise in the context of all of our businesses," Steve Griffiths senior vice president and chief operating officer of Optum Enterprise Analytics, told Business Insider. "So we work collectively with each of our business lines to understand the intelligence within various products." The idea is that by applying artificial intelligence to the massive amounts of data that comes from all of Optum's businesses, it can predict when someone might get sick and solve problems healthcare experts can't on their own.

Despite robotics and artificial intelligence being labelled as threats to the profession, the head of digital legal services for a global firm has downplayed concerns. Speaking ahead of the Lawtech Summit, Tae Royle, head of digital legal services for Ashurst, said reports about the robot lawyer takeover have escalated unnecessarily. "The doom and gloom scenarios are a bit overcooked; the majority of the workforce will adapt rapidly and are getting on the front foot," he said. Mr Royle added that these fears are unsubstantiated at present, as the application of AI remains "very narrow". However, Mr Royle also said that opportunities for AI are "progressively widening, and in five years' time the use cases will be broader and the impacts will be very deep", though he noted there is no need for firms to panic over this, and instead they should properly prepare.