We introduce Confucius3-Math, an open-source large language model with 14B parameters that (1) runs efficiently on a single consumer-grade GPU; (2) achieves SOTA performances on a range of mathematical reasoning tasks, outperforming many models with significantly larger sizes. In particular, as part of our mission to enhancing education and knowledge dissemination with AI, Confucius3-Math is specifically committed to mathematics learning for Chinese K-12 students and educators. Built via post-training with large-scale reinforcement learning (RL), Confucius3-Math aligns with national curriculum and excels at solving main-stream Chinese K-12 mathematical problems with low cost. In this report we share our development recipe, the challenges we encounter and the techniques we develop to overcome them. In particular, we introduce three technical innovations: Targeted Entropy Regularization, Recent Sample Recovery and Policy-Specific Hardness Weighting. These innovations encompass a new entropy regularization, a novel data scheduling policy, and an improved group-relative advantage estimator. Collectively, they significantly stabilize the RL training, improve data efficiency, and boost performance. Our work demonstrates the feasibility of building strong reasoning models in a particular domain at low cost. We open-source our model and code at https://github.com/netease-youdao/Confucius3-Math.

Automated penetration testing (AutoPT) powered by large language models (LLMs) has gained attention for its ability to automate ethical hacking processes and identify vulnerabilities in target systems by leveraging the inherent knowledge of LLMs. However, existing LLM-based AutoPT frameworks often underperform compared to human experts in challenging tasks for several reasons: the imbalanced knowledge used in LLM training, short-sightedness in the planning process, and hallucinations during command generation. Moreover, the trial-and-error nature of the PT process is constrained by existing frameworks lacking mechanisms to learn from previous failures, restricting adaptive improvement of PT strategies. To address these limitations, we propose a knowledge-informed, self-reflective PT framework powered by LLMs, called RefPentester. This AutoPT framework is designed to assist human operators in identifying the current stage of the PT process, selecting appropriate tactics and techniques for each stage, choosing suggested actions, providing step-by-step operational guidance, and reflecting on and learning from previous failed operations. We also modeled the PT process as a seven-state Stage Machine to integrate the proposed framework effectively. The evaluation shows that RefPentester can successfully reveal credentials on Hack The Box's Sau machine, outperforming the baseline GPT-4o model by 16.7%. Across PT stages, RefPentester also demonstrates superior success rates on PT stage transitions.

'The CyberGuy' Kurt Knutsson joins'Fox & Friends Weekend' to discuss the potential effects of artificial intelligence software like ChatGPT on the brain. Using ChatGPT on a long-term basis could have negative effects on brain function. That's according to a study led by the Massachusetts Institute of Technology (MIT), which found that using a large language model (LLM) to write multiple essays over a four-month period could hamper cognitive abilities. In the study, 54 participants were divided into three groups. One group used ChatGPT, an LLM product made by OpenAI, to write an essay.

Meta's chief technology officer has called it "the great honor of my life" to be enlisted in a new US army corps that defence chiefs set up to better integrate military and tech industry expertise, including senior figures from top tech firms that also include Palantir and OpenAI. Andrew Bosworth, a long-term lieutenant to Mark Zuckerberg known widely as "Boz", is one of several senior Silicon Valley executives commissioned to the rank of lieutenant colonel in the corps, called Detachment 201, which the US army says will "fuse cutting-edge tech expertise with military innovation". Bosworth, who joined Facebook in 2006, was sworn into the army reserves earlier this month alongside Shyam Sankar, the chief technology officer of Palantir, a technology firm with extensive defence contracts, Kevin Weil, chief product officer of OpenAI, and Bob McGrew, an adviser at Thinking Machines Lab, a 10bn AI company. They wore military fatigues at the swearing-in ceremony but will not be full-time soldiers. The recruitment is a sign of the increasing importance of technology in modern warfare and growing commercial and research links between some of the largest tech firms and the military.

In an experiment last year at the Massachusetts Institute of Technology, more than fifty students from universities around Boston were split into three groups and asked to write SAT-style essays in response to broad prompts such as "Must our achievements benefit others in order to make us truly happy?" One group was asked to rely on only their own brains to write the essays. A second was given access to Google Search to look up relevant information. The third was allowed to use ChatGPT, the artificial-intelligence large language model (L.L.M.) that can generate full passages or essays in response to user queries. As students from all three groups completed the tasks, they wore a headset embedded with electrodes in order to measure their brain activity.

"We haven't designed or validated AlphaGenome for personal genome prediction, a known challenge for AI models," Google said in a statement. Underlying the AI system is the so-called transformer architecture invented at Google that also powers large language models like GPT-4. This one was trained on troves of experimental data produced by public scientific projects. Lareau says the system will not broadly change how his lab works day to day but could permit new types of research. For instance, sometimes doctors encounter patients with ultra-rare cancers, bristling with unfamiliar mutations.

Teenagers' language might make online bullying hard to detect Generation Alpha's internet lingo is mutating faster than teachers, parents and AI models can keep up – potentially exposing youngsters to bullying and grooming that trusted adults and AI-based safety systems simply can't see. Manisha Mehta, a 14-year-old student at Warren E Hyde Middle School in Cupertino, California, and Fausto Giunchiglia at the University of Trento, Italy, collated 100 expressions and phrases popular with Generation Alpha – those born between 2010 and 2025 – from popular gaming, social media and video platforms. The pair then asked 24 volunteers aged between 11 and 14, who were Mehta's classmates, to analyse the phrases alongside context-specific screenshots. The volunteers explained whether they understood the phrases, in what context they were being used and if that use carried any potential safety concerns or harmful interpretations. They also asked parents, professional moderators and four AI models – GPT-4, Claude, Gemini and Llama 3 – to do the same.

That's why we've created the AI Hype Index--a simple, at-a-glance summary of everything you need to know about the state of the industry. AI agents might be the toast of the AI industry, but they're still not that reliable. That's why Yoshua Bengio, one of the world's leading AI experts, is creating his own nonprofit dedicated to guarding against deceptive agents. Not only can they mislead you, but new research suggests that the weaker an AI model powering an agent is, the less likely it is to be able to negotiate you a good deal online. Elsewhere, OpenAI has inked a deal with toymaker Mattel to develop "age-appropriate" AI-infused products.

Training large language models (LLMs) at scale requires parallel execution across thousands of devices, incurring enormous computational costs. Yet, these costly distributed trainings are rarely verified, leaving them prone to silent errors and potentially wasting millions of GPU hours. We introduce TrainVerify, a system for verifiable distributed training of LLMs. Given a deep learning model's logical specification as the ground truth, TrainVerify formally verifies that a distributed parallel execution plan is mathematically equivalent to it. Direct verification is notoriously difficult due to the sheer scale of LLMs which often involves billions of variables and highly intricate computation graphs. Therefore, TrainVerify introduces shape-reduction techniques and a stage-wise parallel verification algorithm that significantly reduces complexity while preserving formal correctness. TrainVerify scales to frontier LLMs, including the successful verification of the Llama3 (405B) and DeepSeek-V3 (671B) training plans.

Heuristic algorithms play a vital role in solving combinatorial optimization (CO) problems, yet traditional designs depend heavily on manual expertise and struggle to generalize across diverse instances. We introduce \textbf{HeurAgenix}, a two-stage hyper-heuristic framework powered by large language models (LLMs) that first evolves heuristics and then selects among them automatically. In the heuristic evolution phase, HeurAgenix leverages an LLM to compare seed heuristic solutions with higher-quality solutions and extract reusable evolution strategies. During problem solving, it dynamically picks the most promising heuristic for each problem state, guided by the LLM's perception ability. For flexibility, this selector can be either a state-of-the-art LLM or a fine-tuned lightweight model with lower inference cost. To mitigate the scarcity of reliable supervision caused by CO complexity, we fine-tune the lightweight heuristic selector with a dual-reward mechanism that jointly exploits singals from selection preferences and state perception, enabling robust selection under noisy annotations. Extensive experiments on canonical benchmarks show that HeurAgenix not only outperforms existing LLM-based hyper-heuristics but also matches or exceeds specialized solvers. Code is available at https://github.com/microsoft/HeurAgenix.