<p>A Chinese AI startup, Moonshot, has disrupted expectations in artificial intelligence development after its Kimi K2 Thinking model surpassed OpenAI&#8217;s GPT-5 and Anthropic&#8217;s Claude Sonnet 4.5 across multiple performance benchmarks, sparking renewed debate about whether America&#8217;s AI dominance is being challenged by cost-efficient Chinese innovation. Beijing-based Moonshot AI, valued at US$3.3 billion and backed by [&#8230;]</p>
<p>The post <a href="https://www.artificialintelligence-news.com/news/moonshot-ai-gpt-5-claude-comparison-china-breakthrough/">Chinese AI startup Moonshot outperforms GPT-5 and Claude Sonnet 4.5: What you need to know</a> appeared first on <a href="https://www.artificialintelligence-news.com">AI News</a>.</p>

نجحت شركة Moonshot الصينية الناشئة في تجاوز توقعات تطوير الذكاء الاصطناعي بعد أن تفوق نموذجها Kimi K2 Thinking على GPT-5 من OpenAI وClaude Sonnet 4.5 من Anthropic في عدة معايير أداء، مما أثار نقاشًا متجددًا حول ما إذا كانت هيمنة أمريكا في الذكاء الاصطناعي تتعرض للتحدي من الابتكارات الصينية ذات التكلفة المنخفضة. وتُقدّر قيمة Moonshot بمبلغ 3.3 مليار دولار.

La startup china Moonshot ha superado las expectativas al superar el GPT-5 de OpenAI y el Claude Sonnet 4.5 de Anthropic con su modelo Kimi K2 Thinking. Este logro plantea interrogantes sobre el futuro del dominio de la IA, especialmente en relación con la competencia entre las innovaciones estadounidenses y chinas. Con una valoración de 3.3 mil millones de dólares, el éxito de Moonshot destaca el potencial de los avances en tecnología de IA de bajo costo provenientes de China.

La startup chinoise Moonshot a fait sensation en surpassant le GPT-5 d'OpenAI et le Claude Sonnet 4.5 d'Anthropic avec son modèle Kimi K2 Thinking. Cet exploit soulève des questions sur l'avenir de la domination de l'IA, notamment en ce qui concerne le paysage concurrentiel entre les innovations américaines et chinoises. Évaluée à 3,3 milliards de dollars, la réussite de Moonshot met en lumière le potentiel des avancées technologiques en IA à faible coût en provenance de Chine.

Chinese AI startup Moonshot has made headlines by surpassing OpenAI's GPT-5 and Anthropic's Claude Sonnet 4.5 with its Kimi K2 Thinking model. This achievement raises questions about the future of AI dominance, particularly regarding the competitive landscape between U.S. and Chinese innovations. Valued at $3.3 billion, Moonshot's success highlights the potential for cost-effective advancements in AI technology from China.

Chinese AI startup Moonshot outperforms GPT-5 and Claude Sonnet 4.5: What you need to know

arXiv:1906.07172v5 Announce Type: replace-cross 
Abstract: Equivariant neural networks are a class of neural networks designed to preserve symmetries inherent in the data. In this paper, we introduce a general method for modifying a neural network to enforce equivariance, a process we refer to as equivarification. We further show that group convolutional neural networks (G-CNNs) arise as a special case of our framework.

Equivariant neural networks and equivarification

arXiv:2311.13745v4 Announce Type: replace-cross 
Abstract: Diffusion models have become the most popular approach to deep generative modeling of images, largely due to their empirical performance and reliability. From a theoretical standpoint, a number of recent works have studied the iteration complexity of sampling, assuming access to an accurate diffusion model. In this work, we focus on understanding the sample complexity of training such a model; how many samples are needed to learn an accurate diffusion model using a sufficiently expressive neural network? Prior work showed bounds polynomial in the dimension, desired Total Variation error, and Wasserstein error. We show an exponential improvement in the dependence on Wasserstein error and depth, along with improved dependencies on other relevant parameters.

Improved Sample Complexity Bounds for Diffusion Model Training

arXiv:2410.22995v2 Announce Type: replace-cross 
Abstract: A hallmark of advanced artificial intelligence is the capacity to progress from passive visual perception to the strategic modification of visual information to facilitate complex reasoning. This advanced capability, however, remains critically underdeveloped in current Large Multi-modal Models (LMMs). The deficiency is often masked by evaluation metrics that prioritize final-answer accuracy, creating an illusion of competence where genuine reasoning is absent. Using the domain of geometric problem-solving as a precise instrument, we probe this issue through tasks that require constructing visual aids. To this end, we introduce \textbf{VisAidMath}, a challenging benchmark, and our novel Three-Layered Funnel Evaluation Framework. This framework moves beyond simple accuracy (ACCU) to scrutinize the generation of valid visual aids (PVA) and the soundness of subsequent reasoning steps (SPRS). Our extensive experiments on state-of-the-art models, including Doubao-Seed-1.6 and o4, reveal a profound ``Reasoning Illusion''. We observe that high surface-level accuracy conceals a catastrophic failure in the models' ability to produce valid visual aids or to reason from them. Our findings expose a fundamental schism between visual perception and logical deduction in modern LMMs. We host an evaluation platform at CodaBench for testing publicly. Homepage: https://nlp2ct.github.io/VisAidMathHomepage/ Evaluation: https://www.codabench.org/competitions/7634/

VisAidMath: Benchmarking Visual-Aided Mathematical Reasoning

arXiv:2508.01249v3 Announce Type: replace-cross 
Abstract: Large Language Model (LLM) agents offer a powerful new paradigm for solving various problems by combining natural language reasoning with the execution of external tools. However, their dynamic and non-transparent behavior introduces critical security risks, particularly in the presence of prompt injection attacks. In this work, we propose a novel insight that treats the agent runtime traces as structured programs with analyzable semantics. Thus, we present AgentArmor, a program analysis framework that converts agent traces into graph intermediate representation-based structured program dependency representations (e.g., CFG, DFG, and PDG) and enforces security policies via a type system. AgentArmor consists of three key components: (1) a graph constructor that reconstructs the agent's runtime traces as graph-based intermediate representations with control and data flow described within; (2) a property registry that attaches security-relevant metadata of interacted tools \& data, and (3) a type system that performs static inference and checking over the intermediate representation. By representing agent behavior as structured programs, AgentArmor enables program analysis for sensitive data flow, trust boundaries, and policy violations. We evaluate AgentArmor on the AgentDojo benchmark, the results show that AgentArmor can reduce the ASR to 3\%, with the utility drop only 1\%.

AgentArmor: Enforcing Program Analysis on Agent Runtime Trace to Defend Against Prompt Injection

arXiv:2511.14691v1 Announce Type: cross 
Abstract: Attention is the brain's ability to selectively focus on a few specific aspects while ignoring irrelevant ones. This biological principle inspired the attention mechanism in modern Transformers. Transformers now underpin large language models (LLMs) such as GPT, but at the cost of massive training and inference energy, leading to a large carbon footprint. While brain attention emerges from neural circuits, Transformer attention relies on dot-product similarity to weight elements in the input sequence. Neuromorphic computing, especially spiking neural networks (SNNs), offers a brain-inspired path to energy-efficient intelligence. Despite recent work on attention-based spiking Transformers, the core attention layer remains non-neuromorphic. Current spiking attention (i) relies on dot-product or element-wise similarity suited to floating-point operations, not event-driven spikes; (ii) keeps attention matrices that suffer from the von Neumann bottleneck, limiting in-memory computing; and (iii) still diverges from brain-like computation. To address these issues, we propose the Spiking STDP Transformer (S$^{2}$TDPT), a neuromorphic Transformer that implements self-attention through spike-timing-dependent plasticity (STDP), embedding query--key correlations in synaptic weights. STDP, a core mechanism of memory and learning in the brain and widely studied in neuromorphic devices, naturally enables in-memory computing and supports non-von Neumann hardware. On CIFAR-10 and CIFAR-100, our model achieves 94.35\% and 78.08\% accuracy with only four timesteps and 0.49 mJ on CIFAR-100, an 88.47\% energy reduction compared to a standard ANN Transformer. Grad-CAM shows that the model attends to semantically relevant regions, enhancing interpretability. Overall, S$^{2}$TDPT illustrates how biologically inspired attention can yield energy-efficient, hardware-friendly, and explainable neuromorphic models.

Attention via Synaptic Plasticity is All You Need: A Biologically Inspired Spiking Neuromorphic Transformer

arXiv:2511.08955v2 Announce Type: replace-cross 
Abstract: Simulating microstructure evolution (MicroEvo) is vital for materials design but demands high numerical accuracy, efficiency, and physical fidelity. Although recent studies on deep learning (DL) offer a promising alternative to traditional solvers, the field lacks standardized benchmarks. Existing studies are flawed due to a lack of comparing specialized MicroEvo DL models with state-of-the-art spatio-temporal architectures, an overemphasis on numerical accuracy over physical fidelity, and a failure to analyze error propagation over time. To address these gaps, we introduce MicroEvoEval, the first comprehensive benchmark for image-based microstructure evolution prediction. We evaluate 14 models, encompassing both domain-specific and general-purpose architectures, across four representative MicroEvo tasks with datasets specifically structured for both short- and long-term assessment. Our multi-faceted evaluation framework goes beyond numerical accuracy and computational cost, incorporating a curated set of structure-preserving metrics to assess physical fidelity. Our extensive evaluations yield several key insights. Notably, we find that modern architectures (e.g., VMamba), not only achieve superior long-term stability and physical fidelity but also operate with an order-of-magnitude greater computational efficiency. The results highlight the necessity of holistic evaluation and identify these modern architectures as a highly promising direction for developing efficient and reliable surrogate models in data-driven materials science.

MicroEvoEval: A Systematic Evaluation Framework for Image-Based Microstructure Evolution Prediction

arXiv:2511.08195v2 Announce Type: replace 
Abstract: User interface (UI) programming is a core yet highly complex part of modern software development. Recent advances in visual language models (VLMs) highlight the potential of automatic UI coding, but current approaches face two key limitations: multimodal coding capabilities remain underdeveloped, and single-turn paradigms make little use of iterative visual feedback. We address these challenges with an interactive UI-to-code paradigm that better reflects real-world workflows and raises the upper bound of achievable performance. Under this paradigm, we present UI2Code$^\text{N}$, a visual language model trained through staged pretraining, fine-tuning, and reinforcement learning to achieve foundational improvements in multimodal coding. The model unifies three key capabilities: UI-to-code generation, UI editing, and UI polishing. We further explore test-time scaling for interactive generation, enabling systematic use of multi-turn feedback. Experiments on UI-to-code and UI polishing benchmarks show that UI2Code$^\text{N}$ establishes a new state of the art among open-source models and achieves performance comparable to leading closed-source models such as Claude-4-Sonnet and GPT-5. Our code and models are available at https://github.com/zai-org/UI2Code_N.

برمجة واجهة المستخدم (UI) هي جزء معقد من تطوير البرمجيات الحديثة. تظهر التطورات الأخيرة في نماذج اللغة المرئية (VLMs) إمكانية الترميز التلقائي لواجهة المستخدم، ولكن الأساليب الحالية تواجه قيودًا رئيسية: قدرات الترميز متعددة الوسائط لا تزال غير متطورة، ونماذج التفاعل الفردية لا تستفيد كثيرًا من التغذية الراجعة البصرية التكرارية. نحن نتناول هذه التحديات من خلال نموذج UI2Code^N، وهو نموذج لغة مرئية تم تدريبه من خلال مراحل ما قبل التدريب، والتعديل الدقيق، والتعلم المعزز لتحقيق تحسينات أساسية في الترميز متعدد الوسائط.

La programación de interfaces de usuario (UI) es un aspecto complejo del desarrollo de software. Los avances recientes en modelos de lenguaje visual (VLM) muestran un potencial para la codificación automática de UI, pero los métodos actuales enfrentan limitaciones en capacidades multimodales y retroalimentación iterativa. El modelo UI2Code^N aborda estos problemas mediante un enfoque interactivo de UI a código, mejorando el rendimiento al integrar la generación, edición y pulido de UI. Este modelo se entrena a través de preentrenamiento por etapas, ajuste fino y aprendizaje por refuerzo, con e…

La programmation d'interface utilisateur (UI) est un aspect complexe du développement logiciel. Les avancées récentes dans les modèles de langage visuel (VLM) montrent un potentiel pour le codage automatique d'UI, mais les méthodes existantes rencontrent des limitations en matière de capacités multimodales et de rétroaction itérative. Le modèle UI2Code^N aborde ces problèmes grâce à une approche interactive de l'UI vers le code, améliorant les performances en intégrant la génération, l'édition et le polissage de l'UI. Ce modèle est formé par préentraînement par étapes, ajustement fin et appren…

UI programming is a complex aspect of software development. Recent advancements in visual language models (VLMs) show promise for automatic UI coding, yet existing methods face limitations in multimodal capabilities and iterative feedback. The UI2Code^N model addresses these issues through an interactive UI-to-code approach, enhancing performance by integrating UI generation, editing, and polishing. This model is trained using staged pretraining, fine-tuning, and reinforcement learning, aiming to improve multimodal coding significantly.

UI2Code^N: A Visual Language Model for Test-Time Scalable Interactive UI-to-Code Generation

arXiv:2511.11407v1 Announce Type: new 
Abstract: Multimodal Large Language Models are increasingly applied to biomedical imaging, yet scientific reasoning for microscopy remains limited by the scarcity of large-scale, high-quality training data. We introduce MicroVQA++, a three-stage, large-scale and high-quality microscopy VQA corpus derived from the BIOMEDICA archive. Stage one bootstraps supervision from expert-validated figure-caption pairs sourced from peer-reviewed articles. Stage two applies HiCQA-Graph, a novel heterogeneous graph over images, captions, and QAs that fuses NLI-based textual entailment, CLIP-based vision-language alignment, and agent signals to identify and filter inconsistent samples. Stage three uses a MultiModal Large Language Model (MLLM) agent to generate multiple-choice questions (MCQ) followed by human screening. The resulting release comprises a large training split and a human-checked test split whose Bloom's level hard-sample distribution exceeds the MicroVQA benchmark. Our work delivers (i) a quality-controlled dataset that couples expert literature with graph-based filtering and human refinement; (ii) HiCQA-Graph, the first graph that jointly models (image, caption, QA) for cross-modal consistency filtering; (iii) evidence that careful data construction enables 4B-scale MLLMs to reach competitive microscopy reasoning performance (e.g., GPT-5) and achieve state-of-the-art performance among open-source MLLMs. Code and dataset will be released after the review process concludes.

MicroVQA++ هو مجموعة بيانات جديدة عالية الجودة للسببية في علم المجهر مصممة لنماذج اللغة متعددة الوسائط (MLLM). تم اشتقاقها من أرشيف BIOMEDICA وتتكون من عملية من ثلاث مراحل تشمل أزواج الشكل-التعليق المعتمدة من قبل الخبراء، ورسم بياني غير متجانس مبتكر لتصفية العينات غير المتسقة، وأسئلة متعددة الخيارات تم التحقق منها بواسطة البشر. تهدف هذه المجموعة إلى تحسين التفكير العلمي في التصوير الطبي الحيوي، مع معالجة القيود الحالية بسبب نقص بيانات التدريب على نطاق واسع.

MicroVQA++ es un nuevo conjunto de datos de razonamiento en microscopía de alta calidad diseñado para modelos de lenguaje multimodales (MLLM). Se deriva del archivo BIOMEDICA y consta de un proceso en tres etapas que incluye pares figura-leyenda validados por expertos, un novedoso grafo heterogéneo para filtrar muestras inconsistentes y preguntas de opción múltiple revisadas por humanos. Este conjunto de datos busca mejorar el razonamiento científico en imágenes biomédicas, abordando las limitaciones actuales debido a la escasez de datos de entrenamiento a gran escala.

MicroVQA++ est un nouvel ensemble de données de raisonnement en microscopie de haute qualité conçu pour les modèles de langage multimodaux (MLLM). Il est dérivé de l'archive BIOMEDICA et comprend un processus en trois étapes qui inclut des paires figure-légende validées par des experts, un graphe hétérogène novateur pour filtrer les échantillons incohérents et des questions à choix multiples vérifiées par des humains. Cet ensemble de données vise à améliorer le raisonnement scientifique en imagerie biomédicale, répondant aux limitations actuelles dues au manque de données d'entraînement à gran…

MicroVQA++ is a newly introduced high-quality microscopy reasoning dataset designed for multimodal large language models (MLLMs). It is derived from the BIOMEDICA archive and consists of a three-stage process that includes expert-validated figure-caption pairs, a novel heterogeneous graph for filtering inconsistent samples, and human-checked multiple-choice questions. This dataset aims to enhance scientific reasoning in biomedical imaging, addressing the current limitations due to the lack of large-scale training data.

Chinese AI startup Moonshot outperforms GPT-5 and Claude Sonnet 4.5: What you need to know

Was this article worth reading? Share it