AI 论文与研究

A central obstacle in realizing color center cavity-QED-based quantum technologies is inhomogeneity, or disorder, among individual emitters. As the number of emitters increases, so does the complexity of the experimental data, making it di…

Transient thermal fields during solidification connect casting process conditions and cooling behaviour to the formation of defect-prone regions in aluminum cast ings. Repeatedly solving this thermal problem across geometries and process c…

Inference-time scaling enhances a model's reasoning by extending its chain-of-thought (CoT). However, existing approaches typically rely on a single policy trained with outcome-reward reinforcement learning (RL), which often suffers from l…

The MetaFine Open Competition (RoboFineMani workshop, CoRL 2026) evaluates fine-grained manipulation beyond binary success. Its five simulated tasks, from grasping a bottle by an instructed part to inserting letter-shaped pegs with 4 mm cl…

Electrocardiograms (ECGs) may contain physiological information related to hereditary cancer susceptibility, but its relationship with future cancer risk remains unclear. Here, we present ECG2Gene, a two-stage deep-learning framework that …

State of the art latent dynamics world models achieve impressive performance for robot planning. But do they actually encode the physics of the world? In this paper, we show that popular world models today achieve strong results on tasks t…

Natural-language driven smart home control remains challenging: cloud-hosted Large Language Models (LLMs) raise privacy concerns, while local Small Language Models (SLMs) can generate invalid actions or misinterpret requests. We introduce …

Vision-language model (VLM) unlearning is often treated as an objective-design problem: given multimodal forget data, the goal is to optimize a loss that removes undesirable behavior while preserving retained capabilities. We argue that th…

Attribution graphs over the features of sparse autoencoders (SAEs) are used to predict what an ablation does and are compared across training seeds. A TopK SAE keeps the largest preactivations at each token, and a feature's margin is its p…

Conversational text-to-SQL interfaces could give healthcare professionals natural-language access to electronic health record (EHR) data, but evidence for their real-world usability remains limited: prior evaluations have relied largely on…

Sparse autoencoders (SAEs) have recovered interpretable, causally relevant features in robot foundation models (RFMs) such as $π_{0.5}$ (arXiv:2603.19183, arXiv:2603.19233). SAE features on individual instruction tokens are less interpreta…

Does eliciting five weighted estimates within a single response change how a language model responds to information? Across anchoring, hindsight, base-rate, conjunction, and status-quo/default tasks, we compare direct numerical answers wit…

Deploying generative world models for end-to-end closed-loop evaluation and policy learning requires precise scenario control, consistent observations, and sustained generation quality during extended interactions. This report presents \em…

A growing body of work reports benefits from orthogonality constraints in fine-tuning large language models (LLMs). These results raise an open question: do orthogonality constraints also help LLM pre-training? We conduct a systematic empi…

Finding the passage that answers a question is the first step of most practical systems built on documents, and for Tajik we found no openly described model for it. We present Saidzoda-Tajik-E5-Large, a Tajik text embedder built from multi…

Anthropic 发布新研究，用数千个 Claude 智能体自动化红队测试，任务是绕过训练数据监控器。这些智能体生成了表面无害（如摔跤导师对话、养蜂问答）但实际会教出失调行为的训练数据。此类样本可直接用于改进未来的训练数据监控器，作者称这只是第一步，更好的模型和更多算力有望进一步改进红队测试。

Ryan Madden 让 GPT-5.6 Sol 在 Codex 中以低思考级别完成和平模式 Factorio 通关，用时 13.5 小时游戏时间并发射火箭。

Wiz 研究团队梳理 10 月内至少九个 AI 开发者平台披露的安全事件，包括 Fireworks AI、Lightning AI、Railway、Warp、OpenCode 等，多起报告指向凭据窃取、agent swarm 攻击和以“蒸馏”为目的的窃取。

卡内基梅隆大学新论文提出，预训练、微调与RL可共享同一训练规则，将三者数据按任意比例混合往往优于分阶段训练。作者把一种名为MaxRL的RL方法重构为对模型自身正确答案的似然训练，从而让单一训练器混合预训练文本、精选数据和RL样本。

Oxford 博士生、Exin Therapeutics 联合创始人 Gabriel Ocana-Santero 在 X 上描述，Claude Sonnet 5 智能体从 19 种候选基因疗法中自主选择药物、脑区和注射坐标，通过定制电动手术设备为癫痫小鼠完成注射，术中做出 32 次观察，小鼠恢复良好。

MIT 一项新论文调查了来自近 60 所机构的 268 名数学研究生和博士后，76.3% 预计 AI 会让终身教职岗位更难获得，近 70% 的受访者持此看法。调查于 2026 年 9 月和 10 月初完成，多数回复早于 OpenAI 发布 700+ 道开放问题的 AI 解答。在研究中使用 AI 的受访者里，54.8% 表示不喜欢这样做，部分青年数学家因担心被 AI 抢先而减少分享未完成工作。

VoiceArena 推出 Turing TTS 评测，让听众听同一句话的人类语音和模型语音并挑出真人，挑错比例即 Fool Rate。最佳模型 Gemini 3.8 Flash TTS 的 Fool Rate 仅约 10%，而偏好榜前四名模型统计上并列，说明整个领域距难以与真人区分还有距离。

MIT 研究人员构建了名为 LLM Election Observatory 的仪表盘，实时追踪主流 AI 聊天机器人对 2026 年美国中期选举相关问题的回答。团队每隔几天用同一组固定问题重新测试各模型，监测回答如何随选举临近、用户所在地、党派倾向和身份等因素变化。该项目旨在系统记录这一信息环境并向其他研究者和公众开放数据，关注点不止于事实准确性，还包括框架、强调重点、不确定性和引用来源的差异。

UCSD Hao AI Lab 推出 JetSpec++，用因果并行草稿、隐状态预测器和锚点 token 解码器，在当前块验证前先草拟下一块，把草稿与验证流水线化，几乎隐藏全部草稿开销。

Unified Multimodal Models (UMMs) achieve remarkable success on diverse multimodal understanding and generation tasks, yet they still struggle to effectively leverage language-based chain-of-thought reasoning for image generation. As a resu…

Decision models such as Jev have emerged as a highly promising paradigm for LLM agent systems, providing lightweight and reliable categorical reasoning for rapid System One execution. However, existing generation-oriented inference engines…

Agentic workflows are increasingly shifting towards long-horizon tasks, such as open-ended scientific work, where models benefit from long context windows. However, at ultra-long context (above 1M tokens), full attention is prohibitively e…

Planning with world models enables test-time adaptation for embodied agents, offering superior generalization over direct policy learning. However, long-horizon tasks remain hindered by cumulative errors and search complexity. Existing hie…

Recent work shows that a frozen robot policy performs far better inside an agent harness, where control code decides when to query the policy, consult vision tools, or command the arm directly. Yet today these harnesses are built by hand a…

In many practical deployments of Differentially Private Stochastic Gradient Descent (DP-SGD), the server processes data in batches. Randomly sampling these batches allows for better privacy bounds via privacy amplification results. Most sa…

Distribution Matching Distillation (DMD) trains a few-step student using the difference between a teacher's score and an online critic's estimate of the student distribution's score. We investigate what the critic needs, in capacity and in…

Instance-dependent limits for bandit regret, sequential tests, best-arm identification and Markov decision processes are proved by one chain. It runs from a divergence decomposition through data processing to a statistic, a testing step th…

Decision models target structured tasks by directly predicting choices or scores, avoiding autoregressive decoding latency. However, transferring a generative LLM's ranking knowledge to a decision model remains challenging: the teacher exp…

Objective: To distinguish the transferable predictive value of physiological trajectories from that of observation patterns in ICU mortality prediction, and to assess whether measurement timing affects performance independently of observat…

Understanding how gradient descent learns features remains a central challenge in neural network theory. We study this question in a minimal multi-index setting that already exhibits rich multi-phase gradient dynamics: a well-specified two…

In the random allocation sampling scheme each user's data is used in $k$ steps chosen randomly and uniformly from a sequence of $t$ steps. This scheme is known to improve on Poisson sampling in private optimization and in communication-eff…

In real-world deployments, large language model (LLM) systems are increasingly evolving from single-model invocation toward heterogeneous deployments in which different models are dynamically invoked for inference. When the same context is…

The quadratic scaling of attention is a fundamental bottleneck limiting the long-context capabilities of large language models (LLMs). A variety of approaches have attempted to circumvent this, but they either fail to match attention's exp…

Atmospheric fronts are associated with much of the high-impact weather in the mid-latitudes, yet their identification and classification remain largely manual and subjective. Existing deep learning approaches rely primarily on gridded rean…

Distributionally robust optimization (DRO) minimizes worst-case risk under distributional uncertainty, but its decisions can be suboptimal in terms of their excess risk, or regret. We study this gap for multivariate least-squares regressio…