AI Hot Radar · 精选资讯 经过 AI 五维评分精选的 AI 行业资讯 https://github.com/zenitlab/ai-hot-radar zh-CN Thu, 30 Jul 2026 20:00:02 GMT Advancing the price-performance frontier with GPT-5.6 https://openai.com/index/advancing-the-price-performance-frontier-with-gpt-5-6 Explore lower GPT‑5.6 pricing for Luna and Terra—and how OpenAI’s more efficient Thu, 30 Jul 2026 10:00:00 GMT rss_openai Introducing Gemini Robotics ER 2 https://blog.google/innovation-and-ai/models-and-research/google-deepmind/gemini-robotics-er-2/ 谷歌发布Gemini Robotics ER 2，提升机器人视频理解等多能力 Thu, 30 Jul 2026 15:00:00 GMT rss_deepmind A fundamental flaw leaves LLMs strikingly vulnerable to attack https://www.technologyreview.com/2026/07/30/1140927/a-fundamental-flaw-leaves-llms-vulnerable-to-attack/ 研究指出大模型存在固有缺陷，难以完全抵御攻击 Thu, 30 Jul 2026 10:15:19 GMT rss_mit_tech OpenAI 部署 GPT-5.6 优化模型自身推理性能，端到端服务成本最多降低 20% https://www.ithome.com/0/983/590.htm OpenAI用GPT-5.6 Sol优化自身推理性能，服务成本最多降20% Thu, 30 Jul 2026 06:06:15 GMT rss_ithome Spline-Based Boundary Representations for Sparse View Reconstruction and Simulation Using Isogeometric Analysis https://arxiv.org/abs/2607.26234 提出FORGE-SIM方法，从稀疏图像直接重建B样条边界表示用于仿真 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cv HeteroPROPMT: A Real-time and Privacy-Preserving Heterogeneous Collaborative Perception Framework https://arxiv.org/abs/2607.26283 提出异构协作感知框架HeteroPROPMT，兼顾实时与隐私 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cv LumaGuide: Distribution Shaping for Training-Free HDR Generation in Diffusion Models https://arxiv.org/abs/2607.26237 提出无需训练的扩散模型HDR生成框架LumaGuide Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cv BG-REAL: A Public Real-Data Anchored Benchmark for Background Manipulation Detection and Localization https://arxiv.org/abs/2607.26232 提出BG-REAL背景篡改检测定位公开真实数据基准 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cv Lag-aware cross-hand alignment for dual-hand action segmentation https://arxiv.org/abs/2607.26215 提出LACA模块解决双手动作分割的跨手时序偏移问题 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cv When Synthetic Users Fail: A Cross-Domain Benchmark of LLM-Simulated Human Survey Responses https://arxiv.org/abs/2607.26348 提出评估LLM模拟人类调查回复有效性的跨域基准框架 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cl TraceCLIP: Recovering Local Semantics from Patch-to-CLS Contributions https://arxiv.org/abs/2607.26107 提出TraceCLIP，从CLIP的patch到CLS贡献中恢复局部语义 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cv Where Physics Meets Privacy: Federated PINNs for Privacy-Preserving Brain Tumor Biomechanical Modeling https://arxiv.org/abs/2607.26207 提出联邦物理信息神经网络用于隐私保护脑肿瘤生物力学建模 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cv WildShadowRemover: In-the-Wild Video Shadow Removal via Detail-Preserving Video Diffusion Models https://arxiv.org/abs/2607.26203 提出WildShadowRemover，用扩散模型LoRA微调实现野外视频阴影去除 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cv Knowledge-guided Disentanglement with Atomic Actions for Action Recognition https://arxiv.org/abs/2607.26097 提出知识引导的原子动作解耦方法用于复杂场景动作识别 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cv Rad-JEPA 3D: Radiology Joint-Embedding Predictive Model for 3D Computed Tomography https://arxiv.org/abs/2607.26196 提出Rad-JEPA 3D框架，用于3D CT自监督预训练学习表征 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cv Misalignment Has a Personality: A Big Five Account of Emergent Misalignment https://arxiv.org/abs/2607.26389 提出用大五人格解释语言模型微调后出现的广泛错位现象 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cl (Im)Paired Programming: Coding Agents Improve Productivity but Harm Understanding https://arxiv.org/abs/2607.26375 研究编码智能体提升效率但损害开发者代码理解能力 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cl DVPSFormer: Efficient Online Depth-aware Video Panoptic Segmentation for Autonomous Driving https://arxiv.org/abs/2607.26165 提出DVPSFormer在线架构，高效实现自动驾驶深度感知视频全景分割 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cv Symphony of Bias: Exploring Gender Associations with Musical Instruments in Multimodal LLMs https://arxiv.org/abs/2607.26355 研究多模态大模型中乐器相关的性别偏见，构建Symphony-Bias数据集 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cl AgentGUI: An Interface for Observing and Steering Long-Running AI Agents https://arxiv.org/abs/2607.26300 提出AgentGUI，用于观测和引导多并发长运行AI智能体 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cl Evaluating Prompt Scope and Demonstration Similarity in Local LLM Machine Translation https://arxiv.org/abs/2607.26286 研究本地大语言模型机器翻译的提示范围与示例选择影响 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cl DuplexGen: Adaptive Synthesis of Human-AI Turn-Taking Dialogues https://arxiv.org/abs/2607.26178 提出DuplexGen框架，生成适配场景的自适应人机对话轮次 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cl Steering Instruction Hierarchies at Inference Time https://arxiv.org/abs/2607.26228 提出无训练推理时方法V-Steer修复大模型指令层级违反问题 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cl Do Methods Support the Claims? Intra-Paper Verification for Peer Review https://arxiv.org/abs/2607.26066 提出用LLM辅助同行评审中验证论文方法与主张匹配的新方法 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cl Large-Scale ChatBot Validation Through Customer Digital Twin Simulations https://arxiv.org/abs/2607.26060 提出用数字孪生合成客户代理实现大规模聊天机器人验证方法 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cl Choosing Where and How to Moderate: End-to-End Trade-offs in Filter Placement and Response Rewriting https://arxiv.org/abs/2607.26200 研究内容审核分类器部署时的干预位置与响应改写端到端权衡 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_cl FloDR: An invertible dimensionality reduction method based on a normalising flow https://arxiv.org/abs/2607.26278 提出基于归一化流的可逆降维方法FloDR，保留高维数据信息 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_lg Entity Resolution in Practice: Lessons from a Self-Serve Pipeline https://arxiv.org/abs/2607.26298 提出自服务实体解析系统实践的三条现有文献未覆盖的经验 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_lg Early Verdicts, Better Budgets: Sequential Adaptive Rollout Allocation for Compute-Efficient RLVR https://arxiv.org/abs/2607.26253 提出高效RLVR顺序自适应rollout分配方法，降低计算成本 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_lg Between Gradient and Natural Gradient: A Continuum of LoRA Initializations https://arxiv.org/abs/2607.26247 提出统一LoRA初始化框架，揭示梯度类初始化方法的连续谱系 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_lg Weak-to-Strong On-Policy Distillation https://arxiv.org/abs/2607.26246 提出弱到强的在策略蒸馏新范式，解决前沿无更强教师的问题 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_lg Dynamic Parameterization Is Not Dynamic Inference https://arxiv.org/abs/2607.26192 提出冻结控制器审计原则及FCA实现，澄清动态参数化相关误区 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_lg Meta-Learned Reward Shaping for Reinforcement Learning from Human Feedback https://arxiv.org/abs/2607.26094 提出MeRLa元学习奖励塑形框架优化RLHF对齐效果 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_lg Shared SFT Lessons Across Alignment, Model Organisms, and Toy Models https://arxiv.org/abs/2607.26173 研究对齐、模型生物、玩具模型中SFT经验的跨领域迁移 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_lg Data Fusion and Contrastive Alignment for Unconstrained IR Molecular Structure Elucidation https://arxiv.org/abs/2607.26164 提出融合对比对齐方法实现无约束红外光谱分子结构解析 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_lg Emergent Sparsity in Frozen Random CNN Feature Extractors for Deep Reinforcement Learning https://arxiv.org/abs/2607.26059 发现随机冻结CNN特征提取器在深度强化学习中自发产生稀疏表征 Thu, 30 Jul 2026 04:00:00 GMT rss_arxiv_lg 首个突破90%成功率的桌面操作智能体！从堆模型到拼工程，实在Agent登顶OSWorld双冠 https://www.infoq.cn/article/4hUcQzeCeKm0wqkc4Zdc 实在Agent登顶OSWorld双冠，成首个成功率破90%桌面操作智能体 Thu, 30 Jul 2026 02:33:33 GMT rss_infoq Anthropic：Claude 发现 HAWK 致命缺陷，仅用 60 小时攻破 NIST 后量子签名候选算法 https://www.ithome.com/0/983/396.htm Anthropic的Claude发现HAWK算法缺陷，致其被撤出NIST后量子标准候选 Thu, 30 Jul 2026 01:03:11 GMT rss_ithome How enabling two settings tripled our scores on the ARC-AGI-3 benchmark https://openai.com/index/how-two-settings-tripled-our-arc-agi-3-scores 调整两项API设置提升GPT-5.6在ARC-AGI-3基准的得分与效率 Wed, 29 Jul 2026 15:00:00 GMT rss_openai Cryptanalysis of 7-Round AES via the Algebraic Structure of its S-box - www-cdn.anthropic.com https://news.google.com/rss/articles/CBMilwFBVV95cUxNUFFmdVh3azZEbHZjUUVidXkwdkZQWDVTa3RlVGUtWnlWYUlPdWRXVDJicnI5QzlHT0hSd3d1R2dwaXdUSFdWYTZfNG96Qkxlbms2bWlObFZGVGtESmwtOW80SXpCdC0xZVF3dXVnSEg0elBLWm5DWm9sRDVudnd2WnU4UlRVUXF2YlA5VmhfY0hsMnh0eko0?oc=5 Anthropic发布7轮AES基于S盒代数结构的密码分析研究 Tue, 28 Jul 2026 18:50:35 GMT rss_anthropic Mythos Preview’s Chain of Thought in Discovering the AES Möbius Bridge - www-cdn.anthropic.com https://news.google.com/rss/articles/CBMinAFBVV95cUxQMlg0Smw1a0tyQUFwYnlrTDVtbi1OU3hBYzdHcFVaY2JRN0pEcXI2LXh5NUs3SlY3TFV3Y0pYMDg2MVp6bDdCU0ZKUWVGSEU5NENqZEx5WVNJbmZqNG55cnBFYWktSW1LellremptcVNPMXNmeDBiM3lNMDlfeGdBZjBZU2JBdmU0c1dfbGVCdEpnSGllVWFJR1VXMGY?oc=5 Anthropic Mythos Preview用思维链发现AES Möbius桥 Tue, 28 Jul 2026 17:28:33 GMT rss_anthropic How GPT-5.6 fuses frontier intelligence with frontier efficiency https://openai.com/index/gpt-5-6-frontier-intelligence-efficiency GPT-5.6融合前沿智能与效率，提升多场景AI效能 Wed, 29 Jul 2026 00:00:00 GMT rss_openai Accelerating scientific discovery with ChatGPT for Academic Researchers https://openai.com/index/chatgpt-for-academic-researchers OpenAI向10万学术研究者免费开放先进ChatGPT模型 Wed, 29 Jul 2026 10:00:00 GMT rss_openai We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control. https://blog.google/innovation-and-ai/models-and-research/google-labs/lyria-3-5/ 谷歌发布Lyria 3.5音乐生成模型，提升音乐性、歌词等能力 Wed, 29 Jul 2026 16:00:00 GMT rss_deepmind OpenAI open-sources Codex Security CLI to help developers find and fix vulnerabilities from the command line https://the-decoder.com/openai-open-sources-codex-security-cli-to-help-developers-find-and-fix-vulnerabilities-from-the-command-line/ OpenAI开源Codex Security CLI工具，可自动检测修复代码漏洞 Wed, 29 Jul 2026 11:50:56 GMT rss_the_decoder 问世以来最大更新：MCP 2026-07-28 规范发布，转向“无状态”核心 https://www.ithome.com/0/983/102.htm Anthropic发布MCP 2026-07-28规范，核心转向无状态架构 Wed, 29 Jul 2026 07:25:46 GMT rss_ithome LENS: Adaptive Spatio-Temporal Zooming for Keyframe Sampling in Long-Form Videos https://arxiv.org/abs/2607.25125 提出无训练长视频关键帧采样框架LENS，平衡时空细节与覆盖 Wed, 29 Jul 2026 04:00:00 GMT rss_arxiv_cv OPERA: Offline Policy-guided Expert Routing and Adaptation for Universal Biomedical Image Analysis https://arxiv.org/abs/2607.25108 提出OPERA多智能体集成框架，解决生物医学图像分析部署瓶颈 Wed, 29 Jul 2026 04:00:00 GMT rss_arxiv_cv IMPRINT: Image-Conditioned Query Enrichment for Long-Tail Object Goal Navigation https://arxiv.org/abs/2607.25106 提出IMPRINT框架，用图像增强文本查询提升长尾物体导航效果 Wed, 29 Jul 2026 04:00:00 GMT rss_arxiv_cv MorphUNet: Alpha-Controlled Biometric Transport for Diffusion-Based Face Morphing Attacks https://arxiv.org/abs/2607.25092 提出首个扩散式人脸变形攻击框架MorphUNet Wed, 29 Jul 2026 04:00:00 GMT rss_arxiv_cv