AI 论文与研究

As Speech Large Language Models (speech-LLMs) become increasingly multilingual, Arabic remains significantly underrepresented, highlighting the need for dedicated infrastructure to train and evaluate Arabic speech-LLMs. To address this gap…

Zero-cost proxies rank architectures cheaply, but their reliability varies across search spaces. We introduce CoRA-NAS (COarse Ranking + Anchor-residual), a two-stage framework combining a static ranking prior with low-cost learning-curve …

Large language models generate fluent text that can contain unfaithful claims -- a phenomenon known as hallucination. We present a multi-signal detection pipeline combining fine-tuned DeBERTa-v3 classification, Monte Carlo (MC) Dropout unc…

Many biological discovery problems require experiments to be selected sequentially under constrained budgets. CRISPR screening is a prominent example, as exhaustive perturbation testing is often infeasible and candidate perturbations must …

Recently, a wide range of recommendation algorithms inspired by deep learning techniques have emerged as the performance leaders on several standard recommendation benchmarks. While these algorithms were built on different DL techniques (e…

Recursive self-improvement (RSI) enables AI systems to turn experience and feedback into persistent changes that improve both their capabilities and the process of future improvement. We first use the Headroom-Closed Index (HCI) to reveal …

Continuous glucose monitoring (CGM) provides high-frequency measurements of glucose dynamics and enables short-term glucose forecasting for diabetes management. Although time-series foundation models have shown strong general forecasting a…

A language model normally begins training with random word embeddings: whatever 'banana' means must be learned from training corpora. I implement St. Augustine's picture of word learning, meaning by ostension, for a small masked language m…

Anthropic 发布迄今最详细的威胁情报报告，涵盖针对网络攻击、影响行动、监控、生物学和武器制造的 Claude 滥用企图及发现与处置过程。报告中所有行动均已被 Anthropic 阻断，相关经验用于加强防护，并在适当时与当局和其他 AI 公司共享。报告地址：https://www.anthropic.com/threat-intelligence-report-september-2026

Ethan Mollick 指出，数学界当前的处境是"锯齿状前沿"的后果，也是其他职业未来的预演。数学家的工作不只是做证明，还包括指导学生、维护学术社区、守护学科未来、培养对数学的热爱，这些 AI 做不到。他担心 AI 公司聚焦最显眼的"做证明"会损害这些 AI 无法完成的任务，并预测更多职业将被迫向外界说明：自己的工作不只有 AI 能做的那部分。

最初，Transformer 论文是为了改进 Google Translate 而提出的。9 年后，当 AI 产业建立在其架构之上时，我们从未停止思考翻译这件事。 认识 North Small Translate：一个领先的开放机器翻译模型。

AWS 论文提出 UnitBoost，用任务给定的 unit map 将 worker 输出转为槽值提案，以受限 argmax 组装结果，未填充或不支持的槽位形成显式残差指导下一轮，从而无需生成式 meta-agent 来管理复合 LLM 系统。

Multivariate time series imputation (MTSI) aims to recover missing values in temporal data composed of multiple interdependent variables. This problem is central to real-world applications such as healthcare monitoring, traffic networks, a…

Neural audio codecs are a fundamental component of modern speech generation systems. While recent codecs achieve increasingly low bitrates, reducing frame rate remains challenging, as each token must preserve more information while maintai…

The energy transition is reshaping residential electricity consumption through the increasing adoption of distributed generation, electrified appliances, and demand-response programs. Understanding these evolving behaviors requires access …

We present Vidu S2, which comprises Vidu S2-Avatar, a real-time interactive digital-character model, and Vidu S2-Editing, a real-time video editing model. Moreover, we explore the feasibility of real-time spatial video generation for both …

Domain practitioners understand their business constraints but may lack operations-research expertise or dedicated support. LLM-based optimization agents translate natural-language requirements into models or solver programs that establish…

In this work, we develop an inverse Physics-Informed Neural Network (PINN) framework to infer the dependence of the scrape-off layer (SOL) perpendicular heat conductivity on plasma density and temperature, $κ_\perp(n,T)$. The method combin…

High-dimensional dense text embeddings and large language models face real obstacles in financial-disclosure analysis: context-window limits, hallucination risk, high computational cost, and the arbitrary rotation of vector spaces across i…

This paper presents a novel approach for data-driven self-learning control of highly flexible, modular manufacturing systems. Specifically, we employ a novel framework for model-based reinforcement learning which introduces approximate inv…

When forecasting a firm's future financial performance, alternative data - data collected from non-traditional sources such as consumer transactions, web traffic, and prediction markets - can provide timely signals about firms' operating a…

Identifiability of tensor decompositions is often established through linear-algebraic conditions on the factor families. For nonnegative decompositions, however, positivity provides additional information that is not captured by dimension…

Crash-severity models inform screening, dispatch and site prioritization, yet are deployed without a finite-sample statement of what one prediction means. Off-the-shelf guarantees fail here, because the features that make crash severity di…

Continuous monitoring of water surface elevation across river networks is critical for flood forecasting, water resource management, and understanding the global water cycle. Yet, the scarcity of in situ gauges across much of the globe con…

Boundary representation (B-rep) is the standard format used by modern CAD systems for parametric 3D models. It turns out, the exact same solid can be represented by different B-reps: for example, two engineers using different operations, a…

We present EXYGEN (EXplore Your Graphs ENgine), a framework for knowledge graph (KG) understanding that enables conversational access to KGs at scale. We address two questions in sequence. First, how effectively can LLMs perform text-to-SP…

Early diagnosis of melanoma is critical for improving patient survival rates. However, accurately distinguishing melanoma from other skin lesions remains a significant clinical challenge due to the high visual similarity among lesion types…

Meta 的 Muse 助手将支持自定义语音创建，用户可直接在聊天中通过提示词让 Muse 生成任意语音，生成后可在语音模式中选用。该功能可能由 MAI 模型驱动，但 Meta 也常引入其他实验室模型做基准测试。用户还可为 Muse 智能体开启"工作音效"，听到头像"打字"的声音。

Low-resource multilingual text-to-speech (TTS) systems have expanded language coverage, but their robustness under complex text inputs remains insufficiently diagnosed. Existing evaluations mainly focus on naturalness, speaker similarity, …

Large language model (LLM) training is among the fastest-growing sources of electricity demand in modern data centers, and power availability is a primary bottleneck to continued AI infrastructure growth. Making the power consumption of th…

Generative Marginalization Models (MaMs) have been recently introduced as efficient neural sampling models for any-order autoregressive modelling of discrete distributions. By learning both the marginal and conditional probabilities of a p…

Commercial text-to-image systems silently revise user prompts before generating images, a step users typically cannot disable or even see. Yet, existing audits of cultural bias examine only the final images and treat generation as a single…

Reliable, low-latency perception is crucial for Formula Student Driverless vehicles, yet many existing pipelines rely on deep learning and multi-sensor fusion, often requiring GPU acceleration. This paper presents a lightweight LiDAR-only …

Nvidia 首席科学家 Bill Dally："所有人都在展示的，实际上我在 Hot Chips 上确实看到了一些相关幻灯片，就是 SemiAnalysis InferenceX Benchmark。因为它恰好命中了人们想要的数据" Nvidia 首席科学家在上个月与 Google TPU Fellow Norm Jouppi 的 Compute History Museum 座谈会上发表了上述言论。

自变量推出 X2Streaming-TTS，可在文本 token 到达时即时生成语音，严格零前瞻，单请求 TTFT 中位数 15.8 ms，64 并发请求下低于 120 ms。它用因果承诺机制暂存歧义数字、单位和符号，并以语音状态继承跨片段保留音高与音色；在 6/8 项流式评测中识别错误最低，数字与流式歧义测试 CER 为 0%，质量与所评测的离线基线相当。

生数科技在外滩大会发布面向机器人灵巧操作的自进化通用世界模型 Motus2，将策略、世界模拟与价值评估纳入统一视频-动作模型闭环，呈现 RSI 初步特征。该模型被视为从 L3"在世界中行动"走向 L4"自主世界智能体"的探索，并引入触觉反馈。9 月 12 日起团队将陆续公布更多实验结果与技术细节。

阿里巴巴集团拟领投 AI 训练与基准测试初创公司 UniPat AI 的 3 亿美元融资，本轮估值达 25 亿美元，腾讯控股及红杉等老股东也参与投资。UniPat 由去年在通义 AI 实验室从事后训练分析、数据合成和强化学习工作的李宽创办，为 AI 模型设计贴近真实使用环境的测试与评估场景，基准测试涵盖 AI 编程智能体、浏览器智能体及前沿多模态模型的视觉推理能力。

Salesforce AI 发布论文《Co-Evolving Harnesses and Models： On-Policy Correction Helps Weaker Models Catch Up Where Imitation Fails》。

Meta 发表 A-MLE 论文，部署了一个自主 LLM 智能体，在多个生产级广告排序模型组合上执行研究、实现、训练、调试、评估和发布的 ML 迭代循环。

清华大学-美团数智生活联合研究中心将于2026年9月21日举办学术论坛，主题为「物理世界的AI与大模型：挑战、路径与实践」，邀请高校学者与美团技术专家探讨具身智能、世界模型、智能体等方向及AI在本地生活场景的落地。论坛采用线上直播加清华线下形式，线下报名9月17日24：00截止，仅面向清华师生及其他高校教师和优秀在校生。参会福利联合美团LongCat 2.0送出30份、每份1亿Token体验包。