AI 论文与研究

Multimodal large language models (MLLMs) struggle with spatial, embodied, physical, and temporal reasoning. We hypothesize that these failures reflect a shared deficit in visual transition reasoning, and test whether this capability can se…

Generating realistic human-object interactions (HOI) in complex 3D scenes requires two complementary capabilities: reasoning about interaction feasibility in the environment and synthesizing realistic human-object motion. However, supervis…

推荐阅读！ ExploreNet: Learning Where to Explore in Diffusion GRPO

2005 年我刚读研时，写过一篇关于早期计算机黑客/phreaking/BBS 圈子的论文：黑客往往由好奇心驱动，但他们造出的工具被“Script Kiddies”广泛利用，后者造成了大部分破坏与混乱 总之，关于 AI 黑客……

Arena 推出 Arena Alignment Index 基准，基于 27 个模型的 90K+ 真实智能体会话，测量越权操作（UA）、虚假归因（FA）和虚假完成（DC）三类信号。

Anthropic 发布报告，披露在评估和内部使用中观察到的四类 Claude 非预期行为：利用软件漏洞在服务器上运行命令、误提交敏感表单、绕过 token 或付费限制获取数据、以及用 URL 缩短服务绕过 fetch 工具的 URL 长度限制。

Waymo 新研究显示，清醒司机的致命事故卷入率比整体平均水平低 23.3%。但 Waymo 指出，清醒并不等于安全，道路致死是系统性问题，常由低能见度、疲劳和道路基础设施不佳导致。Waymo Driver 不会处于受损状态，专为应对清醒人类司机仍面临的系统性风险而设计。

最重要的 AI 失败可能并非那些显而易见的。 Microsoft 合作伙伴研究经理 Jennifer Neville 向主持人 Chad Atalla 解释，为什么“令人意外的失败”能揭示人类对智能的期望与 AI 系统实际运作方式之间的分歧。https://msft.it/6017aU8r7

社区项目 ReJev 用 LoRA 后训练将面壁开源 2B 模型 MiniCPM5-2B 适配为决策模型，输入状态、问题与候选选项后直接输出一个决策。在其 1,892 条密封留出集上，准确率从 51.11% 提升至 80.50%（+29.39 个百分点），报告评测中无效输出为 0%，累计 Modal 账单约 5.31 美元。

Zenity Labs 发现 AWS Bedrock AgentCore 的漏洞链“AgentCorruption”，只需对一个公开 Agent 发送一条提示词，即可窃取凭证并接管同 AWS 账户和区域内所有 Agent，读取私有对话、下载源代码并篡改长期记忆。

Memory-augmented mobile GUI agents store successful execution trajectories and reuse them in later tasks, but a stored trajectory rarely matches a new task exactly. The new task may use different parameters, share only some of its steps wi…

Self-blended images are widely used to train face-swap detectors, but primarily capture blending artifacts. We investigate whether adding illumination inconsistencies improves detection. Temporal Self-Blended Images (T-SBI) transfer lighti…

Complete orthogonal predictive coordinates do not by themselves bind a latent direction to a named intervention. We present a mathematical and synthetic audit motivated by vascular device-vessel suitcordance. Capacity-matched least-squares…

Unsupervised learning is one of the main branches of machine learning. Here I argue that unlike the other branches of machine learning (supervised and reinforcement learning), unsupervised learning is a rather heterogenous field that can s…

Puzzles are widely used to evaluate the reasoning capabilities of artificial intelligence (AI) systems for sequential decision making, yet approaches originating from different paradigms are rarely compared under unified conditions. To add…

This study addresses the problem of structured sentiment analysis, whose goal is to obtain a fine-grained sentiment graph where the nodes represent spans of sentiment holders, targets, and expressions, while the arcs define the relationshi…

Existing reaction models primarily learn molecular transformations, whereas enzy- matic reactions depend jointly on molecular structure and catalytic function. We formulate this problem as learning an enzymatic reaction space linking react…

Foundation models supply reinforcement learning (RL) with priors that mitigate its longstanding weaknesses in sample efficiency and transfer, but their token-by-token generation makes queries sequential and costly. Jev, a recently released…

In the graph model for conflict resolution (GMCR), a decision maker (DM) either moves the conflict to another state or does nothing. The basic definitions leave inaction implicit, so every action that leaves the state unchanged is treated …

Image-to-3D generation has become increasingly capable of producing objects that closely resemble the input image, and an outstanding challenge is to reproduce the depicted object itself, including the specific geometry that defines it. In…

Fusing a learned camera-noise fingerprint with an RGB appearance backbone is an appealing route to generator-independent deepfake detection, because the noise residual is grounded in image-formation physics rather than in the texture stati…

Verifier scores now serve as both benchmark metrics and training rewards for large language model (LLM) agents, and a change in score is routinely read as a change in capability. It may instead reflect a change in the evaluation. We introd…

Diffusion models generalize early in training and later reproduce individual training samples. Standard tests detect memorization only once one-shot generation produces near-copies, leaving a released model unaudited until its outputs fail…

Reliable predictive uncertainty in Laplace approximations depends critically on the prior precision, yet selecting it requires a posterior sweep that is prohibitively expensive for neural networks with billions of parameters. Under the Max…

Online platforms contain growing numbers of bots, deceptive reviewers, and scam accounts that imitate legitimate users. Such camouflage blurs graph neighborhoods and behavioral attributes, making it difficult for graph neural networks (GNN…

Deep reinforcement learning agents reach strong performance in real-time strategy games but can be brittle against opponents outside their training distribution. Separating strategic command selection from learned unit control allows diffe…

Autonomous driving capabilities depend strongly on the distribution of scenarios encountered during training. Existing methods commonly construct or dynamically adapt training scenario distributions using surrogate criteria such as realism…

On-policy distillation (OPD) supervises student-generated trajectories with token-level teacher signals. Its sampled-token variant avoids the cost of full-vocabulary probabilities. Yet we find that training on fewer tokens can outperform f…

Improving a long-horizon LLM agent means evolving the harness around a frozen model or training its weights. We let a self-evolving harness make the system stronger first, then cross seed and evolved harnesses with base and trained weights…

Automated interviewers and conversational agents are increasingly used in research, recruitment, customer service, and education. However, many existing systems rely on fixed question sequences and provide limited context-based personaliza…

NVIDIA 联合普林斯顿大学和马里兰大学提出 PivotOPD，一种面向多轮 LLM 智能体的在线策略蒸馏方法，训练智能体避免最致命的早期错误并在错误发生后恢复。

Waymo 发表两篇研究指出，即使移除酒驾因素，清醒司机夜间致命车祸风险仍为白天的3.1至3.9倍。研究构建暴露重建模型，估算美国50个人口最多城市中清醒（零 BAC）与受酒精影响司机的行驶里程，发现移除酒精相关司机使城市致命事故率平均下降23%，从每1亿英里1.42降至1.10起。

斯坦福论文提出 DeLM，用共享上下文和任务队列取代中央协调智能体，智能体异步认领任务并实时共享发现。

Evolvent AI 发布开源研究环境 RSIGym，让 AI 智能体在统一预算环境内重训练模型并改写 harness。

一篇新论文指出，AI 将数学证明翻译成 Lean 后，通过 Lean 检查并不能说明原始自然语言证明是否正确。论文展示了聊天机器人通过悄悄修正错误，把一个错误的证明变成有效的 Lean 证明；并论证判断一个陈述能否被忠实翻译，可证明地难于停机问题，因此任何 AI 翻译器都无法总是做到。

数学从这里开始将以惊人的速度发展。 在几个小时内，研究人员就找到了新方法/新证明。 这是乘以超长数字的新方法，比之前最佳理论所允许的更加省力。

这很棒。在很多情况下，算力获取而非资金才是这些团队的瓶颈。 我们仍然需要更多学术/非营利方向的资助，一步一步来。

Apple 研究团队提出 Normalizing Trajectory Models（NTM），将扩散模型每个反向步骤建模为带精确似然训练的条件归一化流，结合每步浅层可逆模块与跨轨迹深层并行预测器，可从头训练或从预训练 flow-matching 模型初始化。

Zyphra Research 提出利用 MoE 路由的层内与跨层相关性来优化专家放置，从而降低 all-to-all 通信开销。实验显示，在 8 台 MI300x 节点集群上，EP8 升至 EP32 时 all-to-all 占训练步时间比例从 13% 升至 45%（top-2 路由），top-6 路由下则从 24% 升至 60%。

As we move through the world and carry out everyday tasks, we encounter objects that may become relevant only later. We are capable of recalling where we left something or what was inside a container, even without knowing we would need it …