AI 论文与研究

Anthropic 的 Fable 5.2 让 Kieran Klaassen 在三天内做出融合 Premiere 剪辑、After Effects 特效与 Ableton 混音的 AI 智能体视频编辑器 Everycut，一周后在公司内部演示。这是其测试过的首个让专业人士不读底层代码就能构建生产级工具的模型，能接手整个产品并持续构建、测试和打磨。

Anthropic 与 Diode Computers 合作，让 Claude Fable 5.2 从简单提示词出发完成免提音箱、语音助手中枢和电子宠物三种设备的设计，包括 PCB 布局、CAD 建模、固件编写和 DFM 检查，三块电路板直接由工厂按设计文件制造，原型经数周使用仍正常工作。

Audio deepfake detectors trained with known synthesis attacks can learn generator-specific shortcuts that fail as synthesis systems and acoustic conditions evolve. We investigate an attack-free alternative: a normality model trained exclus…

The CoRL 2026 RoboEval challenge has eight bimanual tasks, and it ranks a policy not only on whether it succeeds but on how: every stage of the task, done safely, efficiently and with both arms in coordination, scored against human demonst…

Aligning two models without paired data requires a distance that both share, yet existing methods use Euclidean or spherical geometry. This is either an arbitrary choice or imposed by their loss function. We show that, under a precise form…

Temporally correlated noise can substantially improve the utility of differentially private optimization, yet candidate mechanisms are commonly designed or evaluated using only partial information about the optimization process, such as wo…

Agentic autoregressive video generation combines semantic planning, sequential generation, and visual verification for controllable synthesis. Growing visual histories enlarge the key-value (KV) cache, increasing memory use and attention c…

On-policy distillation (OPD) trains a student language model on its own responses, guided by a fixed teacher. Generating these responses is expensive, so systems use quantized inference and asynchronous rollouts to reduce the cost. Both, h…

The order in which propagators are executed can substantially affect the performance of constraint programming solvers, yet propagator priorities are typically determined by fixed, manually designed heuristics. In this work, we explore lea…

Pre-training code language models relies on massive code corpora usually streamed over the internet. However, researchers rarely need a multi-terabyte archive in its entirety; instead, their workflow centers on curated subsets: selecting w…

Evaluating general-purpose multimodal large language models (MLLMs) for surgical scene understanding remains limited due to scarce surgical data and fragmented evaluation protocols. We address these challenges through a two-stage evaluatio…

Meta 发布论文《Learnings and Pitfalls from AI-Assisted A/B Experimentation at Scale》，介绍其 ABTest Agent：LLM 只负责选择和排序实验步骤，所有数字都由经过验证的确定性工具计算，报告模板要求每张表格可追溯到代码，知识库仅收录专家审核过的文档。

上海交大 IPADS 实验室联合化学化工学院团队在《自然》发表成果，构建 AI 原生科学实验室 AutoDNA（思源一号），实现自主分子生物学实验操作。系统采用多智能体架构分解实验流程，把仪器能力抽象为可调用的原子服务，硬件支持微米级定位和微升级液体处理，集成基于 OpenHarmony 的控制终端，已在核酸检测、酶促 DNA 合成、癌症检测等场景完成应用验证。

清华大学论文提出 TraceMRT 框架，把隐藏任务改写为主任务的合理子目标来攻击 CoT 监控。在 SHADE-Arena 上对 GPT-5.2 和 Claude-4.5 监控的平均逃逸率从此前最佳攻击的 27.56% 升至 59.94%；配套的贝叶斯轨迹监控框架将检测 AUC 从 0.26 提到 0.91。作者建议用任务特定攻击做红队测试，并对整体轨迹和关键步骤分别打分。

阿里与多所亚洲高校的研究者基于真实编码会话中的用户修改请求构建了 Co-Design Bench，要求智能体改造现有 Web 应用，输入含图片、音频、视频等。十个受测模型中最好的 GPT-6 Astra 平均满足约 70% 的任务要求，但仅 26.4% 的任务完全正确，Claude Opus 5 与 Qwen3.8-Max 分列其后，Grok 4.6 垫底。

据《WIRED》报道及 Exin Therapeutics CEO 于 11 月 5 日在 X 公布的结果，GPT-6.1 Sol 连续三次参与活体小鼠脑部手术中的 Bregma 和 Lambda 解剖标志定位，每次由外科医生纠正后保存图像与测量数据供下次参考。

芝加哥大学数学教授 Frank Calegari 认为，AI 能批量生成证明后，数学人才招聘应更多依赖面对面交流而非书面材料，30 分钟面谈即可区分批量产出低质量成果者与有清晰研究愿景的候选人。他主张评审委员会不应仅凭一篇论文或一项成果录用学生，并有责任培养研究生的数学表达能力。此前康奈尔大学教授 Steven Strogatz 也指出，能解释 AI 生成结果并提出新问题的数学家更受重视。

Salesforce 新论文发现，从大型外部教师模型做蒸馏通常比纯 RL 消耗更多算力，而将多个 RL 训练的领域专家模型合并可节省 42% 算力。研究按每单位训练 FLOP 的留出准确率对比蒸馏与 GRPO，覆盖不同任务难度、模型规模和领域；使用外部后训练教师时 GRPO 普遍胜出，蒸馏还常导致回复变长。

Stanford 论文提出经验性信息设计框架，在 10 个前沿 LLM 上对比不同披露粒度下的多智能体交易、议价与拍卖表现。

Google 新论文发现，与会反驳的 AI 队友练习的学生写出的商业计划优于使用普通 AI 工具的学生。研究基于 Google Research 平台 Vantage，AI 队友由 Gemini 3.1 Pro 驱动，会提出有缺陷的想法、表达分歧并陈述错误说法，但不替学生完成任务。

Change impact analysis (CIA) helps developers identify code entities impacted by their initial changes. While several approaches for CIA were proposed, they still struggle with either low precision or low recall, thereby increasing manual …

Options market makers absorb client order flow and must hedge the resulting inventory risk under trading frictions. We study whether an option hedging reinforcement learning (RL) agent trained on historical market data outperforms three al…

Cyber-physical smart-grid detectors are typically treated as black boxes and validated exclusively on clean, temporally shuffled datasets, which risk overly optimistic performance claims. This work presents a multi-testbed evaluation study…

This paper introduces COUPLED FLOW, a new neural operator learning framework for multi-physics PDE systems that captures the co-evolution of multiple physical processes by modeling their cross-process interactions. Unlike existing neural o…

Data transformation pipelines shape the training distribution in computer vision, but designing them is a computationally expensive problem. We investigate whether large language models can enable efficient discovery by synthesizing execut…

The field of mechanistic interpretability is well-suited to collaborative projects between mathematicians and AI researchers. This work describes one such project. We trained a small transformer to learn the language of the Coxeter group a…

Researchers increasingly estimate political behavior based on race and gender from voter files, where race is imputed from surnames and geography. This only works where surnames suggest group membership, an assumption violated where surnam…

Existing evaluations of AI models on high-finance work focus on retrieving answers from financial documents and producing numerically correct outputs. Real private equity (PE) and investment banking (IB) work also depends on choosing betwe…

Binary decision-making under uncertainty arises across a wide range of applications, including energy systems scheduling, transportation, and workforce planning. While machine learning has been increasingly applied to accelerate or replace…

In Bangla, a main verb is often followed by one of a small set of vector verbs. These verbs lose their own meaning and add a shade of meaning instead, such as completion. For example, likhe phello means that someone finished writing, not t…

Fixed prompt-based language models face challenges in open-domain conversations, where versatile topics require different expert advice or demonstrations to maximize their universal performance. Motivated by the dynamic prompt studies in l…

Detectors of AI-generated text are usually tuned and tested on passages of several hundred tokens, yet many real cases are short. We study score-based detection as a function of passage length. We give a closed form for how the false-posit…

牛津 BOLD 团队论文研究 LLM 智能体的零样本协同（ZSC），在 9 个小型合作博弈中测试基于 GPT-5.2、Gemini 3 Pro 等模型的文本和代码智能体。

Real-world deployment of federated learning (FL) requires resilience against adversarial clients and data privacy. Byzantine fault captures the worst-case behavior of adversarial clients, while differential privacy (DP) provides statistica…

Standard ensemble methods can improve robustness by mitigating individual model stochasticity, but they frequently fail to generalize beyond the training distribution, leaving systems vulnerable to out-of-distribution (OOD) data. We introd…

Pre-training large parametric "Vision-Language-Action" (VLA) models on large demonstration datasets has recently yielded promising results for general-purpose manipulation. However, these models still struggle under even small domain shift…

Human-AI complementarity captures the idea that a human assisted by an AI system can outperform both the human and the AI acting alone. In prediction tasks, this improvement is assessed against ground truth labels produced through a refere…

Large Language Models (LLMs) exhibit approximate recall: they retain coarse associations about past events while losing fidelity on specific details. Past research on LLMs has quantified different models' forecasting ability in real-world …

Much of the automation handed to language-model agents is procedural: each step needs little intelligence, but a long unattended run must follow a skill or SOP stated once at the start. We show that agents drift from such procedures over l…

Assessing climate risks requires projections of daily temperature distributions, including rare and spatially widespread extremes. Climate simulations are costly and biased, while correcting them toward observations risks suppressing inter…