The tension between OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei is the driving force in today's biggest technological revolution. SAN FRANCISCO/NEW YORK - If not for the intense rivalry between Anthropic and OpenAI, the generative AI boom might not have arrived so quickly. In late 2022, OpenAI caught wind that Anthropic was working on an AI-powered chatbot. OpenAI CEO Sam Altman immediately directed employees to fast-track a competing product, four people familiar with the matter said. Two weeks later, the company released ChatGPT, sparking a technological revolution that promises to overhaul the global economy and the way humans interact.

Developing improved predictive models for multi-molecular systems is crucial, as nearly every chemical product used results from a mixture of chemicals. While being a vital part of the industry pipeline, the chemical mixture space remains relatively unexplored by the Machine Learning (ML) community. In this paper, we introduce CheMixHub, a holistic benchmark for molecular mixtures spanning a corpus of 11 chemical mixtures property prediction tasks. With applications ranging from drug delivery formulations to battery electrolytes, CheMixHub currently totals approximately 500k data points gathered and curated from 7 publicly available datasets. We devise various data splitting techniques to assess context-specific generalization and model robustness, providing a foundation for the development of predictive models for chemical mixture properties. Furthermore, we map out the modelling space of deep learning models for chemical mixtures, establishing initial benchmarks for the community. This dataset has the potential to accelerate chemical mixture development, encompassing reformulation, optimization, and discovery.

Google DeepMind is funding research into the potential dangers of situations where millions of different AI agents interact with each other online. According to Rohin Shah, who directs the company's AGI safety and alignment research, the mass-market arrival of agents that can carry out tasks without human oversight and follow instructions given to them by other agents creates a whole new class of risk . In an effort to address this, Google DeepMind--which made agent-based tools a centerpiece of Google I/O last month --has teamed up with several other organizations to announce a $10 million funding pot for researchers to study the behavior of multi-agent systems and come up with ways to prevent unsafe scenarios. Joining Google DeepMind are Schmidt Sciences, a philanthropic foundation set up by Eric and Wendy Schmidt; ARIA, the UK government's moonshot agency; the Cooperative AI foundation, a UK-based nonprofit research outfit; and Google's charitable arm, Google.org. I asked Shah and James Fox, who leads the Science of Trustworthy AI program at Schmidt Sciences, what they hope to achieve with that $10 million.

The development of reasoning capabilities represents a critical frontier in large language models (LLMs) research, where reinforcement learning (RL) and process reward models (PRMs) have emerged as predominant methodological frameworks. Contrary to conventional wisdom, empirical evidence from DeepSeek-R1 demonstrates that pure RL training focused on mathematical problem-solving can progressively enhance reasoning abilities without PRM integration, challenging the perceived necessity of process supervision. In this study, we conduct a systematic investigation of the relationship between RL training and PRM capabilities. Our findings demonstrate that problem-solving proficiency and process supervision capabilities represent complementary dimensions of reasoning that co-evolve synergistically during pure RL training. In particular, current PRMs underperform simple baselines like majority voting when applied to state-of-the-art models such as DeepSeek-R1 and QwQ-32B. To address this limitation, we propose Self-PRM, an introspective framework in which models autonomously evaluate and rerank their generated solutions through self-reward mechanisms. Although Self-PRM consistently improves the accuracy of the benchmark (particularly with larger sample sizes), analysis exposes persistent challenges: The approach exhibits low precision (<10\%) on difficult problems, frequently misclassifying flawed solutions as valid. These analyses underscore the need for combined training with process supervision and continued RL scaling to enhance reward alignment and introspective accuracy. We hope these findings provide actionable insights for building more reliable and self-aware complex reasoning models.

The study of Neural Tangent Kernels (NTKs) in deep learning has drawn increasing attention in recent years. NTKs typically actively change during training and are related to feature learning. In parallel, recent work on Gradient Descent (GD) has found a phenomenon called Edge of Stability (EoS), in which the largest eigenvalue of the NTK oscillates around a value inversely proportional to the step size. However, although follow-up works have explored the underlying mechanism of such eigenvalue behavior in depth, the understanding of the behavior of the NTK during EoS is still missing. This paper examines the dynamics of NTK eigenvectors during EoS in detail. Across different architectures, we observe that larger learning rates cause the leading eigenvectors of the final NTK, as well as the full NTK matrix, to have greater alignment with the training target. We then study the underlying mechanism of this phenomenon and provide a theoretical analysis for a two-layer linear network. Our study enhances the understanding of GD training dynamics in deep learning.

Self-attention has emerged as a fundamental component driving the success of modern transformer architectures, which power large language models and various applications. However, a theoretical understanding of how such models actually work is still under active development. The recent work of (Marion et al., 2025) introduced the so-called single-location regression problem, which can provably be solved by a simplified self-attention layer but not by linear models, thereby demonstrating a striking functional separation. A rigorous analysis of self-attention with softmax for this problem is challenging due to the coupled nature of the model. In the present work, we use ideas from the classical random energy model in statistical physics to analyze softmax self-attention on the single-location problem. Our analysis yields exact analytic expressions for the population risk in terms of the overlaps between the learned model parameters and those of an oracle. Moreover, we derive a detailed description of the gradient descent dynamics for these overlaps and prove that, under broad conditions, the dynamics converge to the unique oracle attractor. Our work not only advances our understanding of self-attention but also provides key theoretical ideas that are likely to find use in further analyses of even more complex transformer architectures.

Many of the insights hitting soccer pitches today trace back to Jesse Davis and a team of computer scientists open-sourcing tools for some of the sport's trickiest problems. Imagine tuning in to the opening kickoff of a World Cup match and seeing a player intentionally send the ball all the way down the pitch and right out of bounds on the opponent's end. Casual fans might scratch their heads. If you were Jesse Davis, though, you'd know that this play could be a prime setup to score. Davis is a professor of computer science at KU Leuven in Belgium and head of its Sports Analytics Lab, which has been at the vanguard of a data awakening in soccer since its inception more than a decade ago. Though the research group brings machine-learning models to bear on a variety of sports--including basketball, volleyball, and field hockey--nowhere is its impact felt more than on the soccer pitch.

Chemist Tim Cernak is using two decades of experience in Big Pharma to try to save Gila monsters, loggerhead sea turtles, and many more creatures. In 2018, after nearly two decades working in Big Pharma, chemist Tim Cernak was ready to put his skills to a new use. For Merck, he'd developed precision therapies for cancer, HIV, and diabetes that could target disease while minimizing harm to healthy cells. But as a lifelong nature lover, he was increasingly concerned about the health of ecosystems and wondered whether his expertise could transfer. Animals, he learned, are often treated with pharmaceuticals formulated for humans, which affect them like old-school cancer drugs: Though intended to kill abnormal cells, they're indiscriminate in the harm they cause. For instance, the standard of care for frogs infected with a deadly skin infection is itraconazole, an antifungal that is often lethal for the amphibian.

Recently, slow-thinking systems like GPT-o1 and DeepSeek-R1 have demonstrated great potential in solving challenging problems through explicit reflection. They significantly outperform the best fast-thinking models, such as GPT-4o, on various math and science benchmarks. However, their multimodal reasoning capabilities remain on par with fast-thinking models. For instance, GPT-o1's performance on benchmarks like MathVista, MathVerse, and MathVision is similar to fast-thinking models. In this paper, we aim to enhance the slow-thinking capabilities of vision-language models using reinforcement learning (without relying on distillation) to advance the state of the art. First, we adapt the GRPO algorithm with a novel technique called Selective Sample Replay (SSR) to address the vanishing advantages problem.

Long reasoning models (LRMs) such as OpenAI's o1 and DeepSeek's R1 have demonstrated remarkable abilities in complex reasoning tasks by scaling test-time compute and exhibiting human-like deep thinking. However, we identify a phenomenon we term underthinking, where LRMs frequently switch between different reasoning thoughts without sufficiently exploring promising paths to reach a correct solution. This behavior leads to inadequate depth of reasoning and decreased performance, particularly on challenging mathematical problems. To systematically analyze this issue, we conduct experiments on three challenging test sets and two representative open-source LRMs, revealing that frequent thought switching correlates with incorrect responses. We introduce a novel metric to quantify underthinking by measuring token efficiency in incorrect answers. To address underthinking, we propose a decoding strategy with thought switching penalty (Tip) that discourages premature transitions between thoughts, encouraging deeper exploration of each reasoning path. Experimental results demonstrate that our approach improves accuracy across challenging datasets without requiring model fine-tuning. Our findings contribute to understanding reasoning inefficiencies in LRMs and offer a practical solution to enhance their problem-solving capabilities.