AI 论文与研究

Elith 与 NVIDIA 发布 Nemotron-Personas-Finance-Japan，这是开源人物数据集 Nemotron-Personas-Japan 的金融领域扩展，据称是首个百万记录级、结合日本官方统计与家庭金融数据的开放人物数据集。

Self-recurrence represents an important component for temporal processing in neural networks, both artificial and biological, where it appears naturally in the form of leaky integration. However, due to the complex temporal dependencies of…

Prior beliefs about whether a virtual human (VH) is controlled by a human or a computer can shape how users interpret and respond to its behaviour. As emerging hybrid VH systems allow control to transfer between human operators and compute…

This work proposes WaveGen, a fully end-to-end diffusion-based text-to-waveform framework based on Spectral Forcing, which jointly learns an internal spectral trajectory and the target waveform trajectory within a single model, enabling di…

"System 1" decision models (pioneered by TypeSafe AI's Jev) replace autoregressive text generation with a single forward pass returning typed probability distributions over declared schemas (Choice, Score, Noul). While deployed for low-lat…

Accurate prediction of electronic Hamiltonians would enable broad property inference while avoiding the high computational cost of Density Functional Theory (DFT). However, progress toward general-purpose materials foundation models is lim…

Generalizing deepfake detection across facial manipulation and fully generated video remains challenging due to forgery traces varying with the manipulation method and scene content. Explicit spatial and temporal feature differences offer …

Does grouping translation updates before speech synthesis change the measured completion gap between two translators? We compare StreamAtt with its fixed-audio-history baseline on ten English-German talks. For each translation, we keep the…

Bayesian optimization is widely used in active learning loops for materials science. Gaussian processes serve as the de-facto surrogate model family, but require task-dependent kernel choices and can lack the modeling capacity for complex …

AI-text detectors reach near-perfect scores in evaluations that contrast fully human-written texts with texts generated from researcher-designed prompts. In authentic student writing, however, AI use is voluntary and often partial. We stud…

Answer accuracy alone does not distinguish failures in acoustic perception, rule knowledge, and rule application. AuDiR (Audio Diagnostic Reasoning) examines these distinctions through controlled input interventions. Each item contains an …

We investigate convolutional-neural-network (CNN) based automatic solver selection for sparse linear systems, a problem of practical importance where solver performance varies dramatically with matrix sparsity structure. A CNN is trained t…

Foundation models can supply predictions for a downstream task that draw on knowledge from extensive pretraining, but they differ in task-specific accuracy and query cost. We develop a framework that improves prediction for the task under …

Speculative decoders raise throughput by banking more target tokens per forward pass, and Medusa, EAGLE-3, DFlash and DSpark all do it the same way: more candidates that survive verification. All of them draft in the target's own vocabular…

Causal discovery from observational data is central to scientific reasoning, yet established constraint-based and score-based methods often recover only a Markov equivalence class and provide no signal in bivariate or strongly autocorrelat…

Symbolic knowledge (motif counts, ring counts, label constraints) is routinely added to graph neural networks (GNNs) and justified by their limited expressivity. We replace this yardstick with an exact one. The largest log-loss reduction a…

Probabilistic circuits (PCs) have emerged as generative models guaranteeing tractable probabilistic inference. However, tractability comes at the price of larger model sizes than in other generative models. As such, to scale PCs to high-di…

UCLA与东北大学论文发现，AI智能体约18%的输出token属可移除浪费，剪掉后任务成功率几乎不受影响。研究覆盖6个前沿模型和4个benchmark，在Kimi-K3的推理中剪除重复与填充内容后，SWE-bench Verified上减少41%输出token，成功率89.5%对89.0%。作者建议按任务与harness定位浪费来源，而非单纯限制token。

Anthropic 的玩具实验中，1000 个运行 Claude Fable 5.2 早期版本的智能体在无 Prove2Me 平台和研究者指导的情况下复现了费马大定理的形式化证明，耗时 28.7 小时，输出 token 仅为原 9 月形式化的三分之一，证明长度为七分之一。

UC Berkeley、中国科学技术大学等机构的论文发现，通过 harness evolution 在模型冻结情况下自改提示词、技能、工具或代码的智能体，平均分提高的同时会丢失原本能完成的任务。

小红书智创 AIGC 团队提出多图合成一致性优化框架 DeCoRL，将整体一致性拆解为目标级判断，通过自动建立目标对应关系、逐个评估身份与外观，再以分解奖励引导强化学习后训练。

MIT 与 NYU 研究者将在本月 CoRL 上发表 FLEX，一个面向机器人运动规划的特征编辑流程，让机器人从自然语言指令中推理出未明确说明的偏好。FLEX 分三步：Add 将用户提到的物体加入特征列表，Generalize 借助 GPT-4o 从已有特征泛化出更高层规则，如除笔记本电脑外也避开其他电子产品；Mask 用视觉系统删除环境中已不存在的物体，整个流程约五秒。

MIT、UBC 等机构的研究者提出统一深度学习系统 WorldParticle，用同一套架构模拟从流动沙粒到蛋白质折叠的多种物理现象。该系统把材料表示为大量粒子，由神经网络逐步预测粒子位置与属性，无需为每种现象手写方程和求解器。团队演示了可变形物体、颗粒材料和蛋白质等六类现象，精度高于数值模拟方法。

Anthropic 发布研究报告，分析 2026 年 8 月 2 日至 9 月 30 日的 20.7 万次科研会话，显示科学家主要将 Claude 用于数据分析和写作，数据收集与假设生成较少。

蚂蚁 inclusionAI 发布 ZoomGain，利用放大后才显现的视觉细节反推整图问题，为视觉语言模型构建监督信号。同步开源 ZoomGainBench（1000 题、941 张图，覆盖属性识别、计数与 OCR）、9000 条 RL 训练数据及 ZoomGain-4B、ZoomGain-9B 模型检查点。

Artificial Analysis 发布 Artificial Analysis Decision Index v2，新增奖励正确性与校准置信度的 Decision Score 头条指标，私有数据集占比从 25% 提升至 45%。

斯坦福与Google DeepMind提出TURTLE方法，用AlphaEvolve对AppWorld等静态工具调用基准的现有任务做变异，加入过滤条件、额外步骤和隐藏查询，把过易的基准改造成仍可解但更难的测试。

Amazon 论文审计 9 个政策类决策基准，发现用于评测模型处理税务、法律和商业规则的基准常把正确答案判错，原因在基准自身的参考答案或评分代码有误。

清华与 Moonshot AI 等机构发布 TCSExam 基准，用 100 道专家评审任务评估 LLM 在理论计算机研究全谱系的能力，涵盖 Lean 定理证明、求解器工程、构造纪录和 30 个公开猜想，所有提交由 Lean 内核或任务检查器验证。

We describe the system submitted by Telefónica Innovación Digital to the Speech-COSER Challenge 2026, which features spontaneous speech from elderly speakers recorded in rural Spain. For ASR, we perform CTC-based pseudo-forced alignment sc…

We study stochastic bandits with reward-dependent delays, possibly infinite, whose reports omit the originating round. For each $c\in[0,1)$, Balanced Historical Optimism (BHO) has a delay term controlled by the largest suboptimal-arm quant…

Predictive algorithms are increasingly used to guide hospital discharge decisions, yet these models can exhibit censoring bias for vulnerable patients due to insufficient historical data or past discharge decisions. When biased readmission…

Many studies of animal vocalizations require clean recordings, but animals are usually recorded in the field, with background noise from weather, vehicles, and other animals. Animal vocalization denoising is challenged by the scarcity of c…

Existing evaluations often test persona expression rather than whether persona information persists in extended decision-making. We introduce THEATRON, a controlled benchmark that uses financial replay to evaluate persona persistence in lo…

We present MicroSim, a system for binary code similarity detection that represents each function as a graph built from IDA Pro Hex-Rays microcode and compares such graphs with a contrastive Siamese graph neural network (GNN). We evaluate o…

In multi-institutional biomedical data analysis, comprehensive disease profiles are often fragmented across different data silos. Furthermore, local data distributions typically exhibit varying levels of granularity or even complex nested …

Swarm Learning improves resilience and removes the need for a single trusted coordinator by distributing model aggregation and validation across participating nodes. However, it also delegates the validation of model updates to peers whose…

The small open embedder EmbeddingGemma-300M, used as released, finds the right passage first for only 51.7% to 60.8% of Belebele questions in Tajik, Kazakh, Uzbek and Kyrgyz. We present Saidzoda-CA-EmbeddingGemma-300M, one model for all fo…

Generating accurate CAD models requires substantial engineering expertise and careful control of geometry, topology, and mechanical constraints. Text-to-CAD aims to simplify this process, but current LLM-based methods often generate CAD pr…

Large language models (LLMs) perform well on general coding tasks, but their capabilities in application-layer embedded software development remain underexplored. Existing embedded benchmarks primarily focus on hardware-in-the-loop validat…