导读：本文围绕“自然20261022出版一周论文导读”，按时间顺序梳理事件经过与各方表态，并标注仍需后续确认的信息。
《自然》（20261022出版）一周论文导读
Science, 22 October 2026, Volume 658, Issue 8137
《自然》2026年10月22日，第658卷，8137期
材料科学Materials Science
Wave coarsening drives time crystallization in mechatronic metamaterials
波粗化驱动机电超材料中的时间晶化
▲ 作者：Jonas Veenstra, Jack Binysh, Vito Seinen, Roger Solé Borràs, Rutger Naber, Damien Robledo-Poisson, et al.
▲链接：
Abstract：When metals are magnetized, emulsions phase separate, or galaxies cluster, domain walls and patterns form and irremediably coarsen over time. This coarsening is universally driven by diffusive relaxation towards equilibrium. Here we discover a vibrational counterpart—wave coarsening—in active elastic media: vibrations emerge and spontaneously grow in wavelength, period and amplitude, before a globally synchronized state called a time crystal forms. We observe wave coarsening in one- and two-dimensional mechatronic metamaterials and capture its dynamical scaling. We further arrest the process by breaking momentum conservation and reveal a far-from-equilibrium nonlinear analogue to chiral topological edge modes. Our results open new questions about the transient physics of systems with non-potential interactions and suggest an organizing principle for nonlinear waves in acoustics, optomechanics, living matter and soft robotics.
Competing mechanisms toughen bcc binary and medium-entropy alloys at 20 K
竞争机制实现体心立方二元中熵合金在20 K下增韧
▲ 作者：David H. Cook, Xiaoyu Chen, Michael O’Leary, Satish Rao, Wenqing Wang, Calvin H. Belcher, et al.
▲链接：
摘要：体心立方（bcc）金属与合金的塑性变形通常由螺型位错主导。其温度敏感运动归因于材料强度随温度降低而急剧升高，从而压缩裂纹尖端塑性区，进而引发韧脆转变。
研究组测试了电子结构相近的IV-V族bcc难熔合金在20K下的力学性能，发现这类材料的拉伸屈服强度与断裂韧性随温度变化呈现平台状特征，不存在明显的韧脆转变。
该行为源于孪生与ω相变的全域激活：这两种机制会先在界面处形核产生位错，之后才发生位错的整体运动。这种初始机制对温度不敏感，并控制屈服过程，因此裂纹尖端塑性区不会随温度降低而收缩。
▲ Abstract：Plastic deformation in body-centred cubic (bcc) metals and alloys is conventionally mediated by screw dislocations. Their temperature-sensitive motion is attributed to a steep rise in strength with decreasing temperature that constricts plastic zones at crack tips, leading to a ductile-to-brittle transition. Here we evaluated the mechanical properties of group IV–V bcc refractory alloys with similar electronic structures at 20?K and found that these materials instead exhibit a plateau in tensile yield strength and fracture toughness as a function of temperature, with no apparent ductile-to-brittle transition. This behaviour arises from the global activation of twinning and an ω-phase transformation, which nucleate dislocations at interfaces before global dislocation motion. The initial mechanism is temperature insensitive and controls yielding, thereby preventing a crack-tip plastic zone from constricting with decreasing temperature.
信息科学Information Science
Retrofitting language models to operate over bytes
改造语言模型实现字节级运算
▲ 作者：Benjamin Minixhofer, Tyler Murray, Tomasz Limisiewicz, Anna Korhonen, Luke Zettlemoyer, Noah A. Smith, et al.
▲链接：
摘要：人工智能（AI）的最新进展很大程度上由大语言模型（以词元为离散处理单元的深度神经网络）来推动。为了表示文本，大多数大语言模型使用单词或单词片段作为词元，这被称为子词分词。
但子词分词会掩盖细粒度信息，这会带来一系列问题，尤其是对于计算机代码或生物序列等依赖单字符或字节承载语义的科学数据很不友好。直接处理文本字节编码的模型避免了这些限制，但性能长期不及基于子词的模型。
研究组提出字节化方法搭建字节级大语言模型，其性能接近基于子词的模型。借助两阶段转换策略，仅需少量额外训练即可把现有基于子词的模型改造成字节级模型。
由此生成的模型优于早期字节级方案，尤其擅长字符级推理任务；通过高效处理字节级信息达到实用推理速度，并通过复用源大语言模型周围的现有生态获得适应性。
该研究工作消除了端到端字节级语言建模的长期性能短板，说明原始文本编码模型在规模化上具备竞争力，在需要细粒度文本理解的场景中更具优势。
▲ Abstract：Recent advances in artificial intelligence (AI) have largely been driven by large language models, deep neural networks that operate over discrete units called tokens. To represent text, most large language models use words or word fragments as the tokens, known as subword tokenization. Subword tokenization obscures fine-grained information, which is problematic, especially for scientific data—such as computer code or biological sequences—where meaning depends on the individual characters or bytes. Models that instead operate directly on the byte encoding of text avoid these limitations, but until now they have lagged behind subword-based models in performance. Here we introduce a general method for creating byte-level large language models through byteification that approach the capabilities of subword-based systems. We use a two-stage conversion procedure to retrofit existing subword-based models into byte-level models with minimal extra training. The resulting models outperform earlier byte-level approaches and excel on character-level reasoning tasks, achieving practical inference speeds by efficiently processing byte-level information and adaptability by reusing the existing ecosystem around the source large language model. Our results remove a long-standing performance barrier to end-to-end byte-level language modelling, demonstrating that models operating on raw text encodings can scale competitively while offering advantages in domains requiring fine-grained textual understanding.
化学Chemistry
Hydrogen bonding in water under extreme confinement
极端受限条件下水中的氢键合
▲ 作者：Xintong Xu, Matthias Kuehne, Harrison A. Walker, De-Liang Bao, Xin Jin, Yu-Ming Tu, et al.
▲链接：
Abstract：Fluids under extreme confinement or near interfaces exhibit molecular structures and intermolecular bonding distinct from their bulk analogues. Optical vibrational studies have shown that interfacial and confined water can exhibit altered intramolecular O–H stretching frequencies, a sensitive spectral signature of changes in intermolecular hydrogen bonding. Investigating confined water experimentally at the length scale of intermolecular and surface forces has, however, remained a challenge. Here we report direct molecular-level observations of hydrogen bonding in water confined inside individual carbon nanotubes (CNTs), enabled by in situ vibrational electron energy loss spectroscopy (vEELS) with nanoscale resolution. Water in larger CNTs exhibits the bonded O–H vibrations of bulk water, but at smaller diameters, and the frequency blueshifts to near the free O–H stretch found in water vapour and water located near hydrophobic surfaces, indicating a highly dispersed, non-H-bonded environment. Theoretical analysis based on quantum vibrational oscillators links the observed spectral features to local hydrogen-bonding configurations, consistent with the experimental observation. Furthermore, cryogenic experiments provide insights into complex structural phase transitions of confined water. This research reveals the quantum and dynamic nature of hydrogen bonds under confinement and the potential impact of unveiling molecular-level structure and bonding in confined fluids.
地球科学Earth Science
Interhemispheric monsoon dynamics over the past 450,000 years
过去45万年间的半球间季风动力学
▲ 作者：E. J. Olson, D. P. Gillikin, D. Parmenter, S. Burns, A. Verheyden, O. Miyawaki, et al.
▲链接：
Abstract：The global monsoon system provides freshwater to much of Earth’s biota, including its human population. The Milankovitch theory predicts changes in Earth’s solar insolation over geologic time, and these variations are expected to result in changes to the global monsoon. Internal climate forcings, such as ice cover and sea surface temperatures, may also affect the system. However, few records of monsoon variability in the Southern Hemisphere over glacial–interglacial cycles exist to test our understanding of the factors that drive variations in the global monsoon. Here we show that monsoons in both the Southern and Northern Hemispheres vary with Milankovitch forcing, and with internal boundary conditions over the past 450,000 years. We present a record of South American Summer Monsoon (SASM) variability, based on speleothem oxygen-isotope data from the Amazon Basin. South American Summer Monsoon variability and that of a major Northern Hemisphere monsoon—that is, the Asian monsoon—have been strikingly out of phase. However, we observe occasional synchronous behaviour in regions impacted by both monsoons. We identify a synchronicity threshold that is exceeded during ten events, including during glacial terminations. Comparison with palaeoclimatic forcing factors during these intervals indicates that greenhouse gas concentrations, cross-equatorial thermal gradients and sea surface temperatures in both the Atlantic and Pacific probably influence these events. This record demonstrates that the global monsoon can reorganize itself, especially during intervals of rapidly changing planetary boundary conditions.
Advancing subseasonal forecasting with machine learning
机器学习助力次季节天气预报
▲ 作者：Hannah Guan, Soukayna Mouatadid, Paulo Orenstein, Judah Cohen, Haiyu Dong, Zekun Ni, et al.
▲链接：
摘要：决策者依靠天气预报开展作物种植、野火管控、水与能源调配以及极端天气防范工作。得益于基于物理的动力学模型与数据驱动人工智能（AI）模型的稳步发展，当前这类预报在两周时效内已达到前所未有的精度。但受误差累积、模型系统性偏差以及大气混沌特性的影响，模型的预报能力在次季节尺度（未来2-6周）会急剧下降。
为解决预报性能下降，研究组提出概率偏差校正，即一种通过学习校正历史概率预报，从而大幅降低系统性误差的机器学习框架。将其应用于欧洲中期天气预报中心（ECMWF）的主流动力学与AI模型后，概率偏差校正使AI预报系统的适度次季节预报技能提升一倍。
对于操作去偏动力学模型，其气压、气温、降水预报目标的提升占比分别达到91%、92%、98%。
研究组为操作部署设计了概率偏差校正，在ECMWF 2025年实时预报竞赛中，基于该方法的全球预报在所有气象要素与预报时效上均排名第一，性能优于6家业务预报中心的动力学模型、一套国际动力学多模型集合、ECMWF AI预报系统以及全球34支队伍的预报系统。
这类概率预报技能提升有助于更精准地预测极端事件，有望助力脆弱地区优化农业规划、能源调度与防灾准备。
▲ Abstract：Decision-makers rely on weather forecasts to plant crops, manage wildfires, allocate water and energy and prepare for weather extremes. Today, such forecasts enjoy unprecedented accuracy as far as 2?weeks thanks to steady advances in physics-based dynamical models and data-driven artificial intelligence models. However, model skill drops precipitously at subseasonal timescales (2–6 weeks ahead) due to compounding errors, systemic model biases and the chaotic nature of the atmosphere. To counter this degradation, we introduce probabilistic bias correction, a machine learning framework that substantially reduces systematic error by learning to correct historical probabilistic forecasts. When applied to the leading dynamical and artificial intelligence models from the European Centre for Medium-Range Weather Forecasts (ECMWF), probabilistic bias correction doubles the modest subseasonal skill of the AI Forecasting System and improves the skill of the operationally debiased dynamical model for 91% of pressure, 92% of temperature and 98% of precipitation targets. We designed probabilistic bias correction for operational deployment, and, in the 2025 real-time forecasting competition of ECMWF, its global forecasts were the best for all weather variables and lead times, outperforming the dynamical models from six operational forecasting centres, an international dynamical multimodel ensemble, the AI Forecasting System of ECMWF and the forecasting systems of 34 teams worldwide. These probabilistic skill gains translate into more accurate prediction of extreme events and have the potential to improve agricultural planning, energy management and disaster preparedness in vulnerable communities.特别声明：本文转载仅仅是出于传播信息的需要，并不意味着代表本网站观点或证实其内容的真实性；如其他媒体、网站或个人从本网站转载使用，须保留本网站注明的“来源”，并自负版权等法律责任；作者如果不希望被转载或者联系转载稿费等事宜，请与我们接洽。