具身智能落地的最后20%，藏在「云」里

今年的WAIC和WRC，具身智能展台的风向变了：不再执着于炫技，转向务实。去年的画风还是十八般武艺，同台竞技——机器人跳街舞、格斗秀、走猫步等；而今年则是走进真实场景：机器人开始被放进物流、工业、家庭等具体任务里。…

• 过去一个多月，雷峰网一线走访了灵初智能、艾欧智能等具身智能公司，也看了各厂商在展台上的演示，一个感受越来越强烈：真实场景里的任务，远比Dem…
• 在走访和交流中，雷峰网也与多位云专家进行了交流。
• 当被问及机器人落地、商业化时，一位云专家表示，具身智能存在明显的“长尾效应”：前60%、70%、80%的进展可能很快，但最后20%却需要花非…

RSS 官方收录 · 可信分层展示

Analysis of Prompt Engineering for Drug Toxicity Prediction

arXiv:2609.03635v1 Announce Type: new Abstract: Clinical trials in the UK can cost up to {\pounds}1.…

• 3 million, with approximately 90% drug failure rate.
• Toxicity is a major contributing factor in drug failure.
• Testing is time and cost intensive.

RSS 官方收录 · 可信分层展示

A computable representation of the physical laboratory enables verifiable workflows

arXiv:2609.…

• 03621v1 Announce Type: new Abstract: Making science computable require…
• A computable representation of the physical laboratory is established …
• It provides the physical-world counterpart to machine-readable knowled…

RSS 官方收录 · 可信分层展示

KC-Bench: A Dynamic Interactive Benchmark for Evaluating Knowledge Conflicts in LLM Agents

arXiv:2609.…

• 03588v1 Announce Type: new Abstract: As LLMs increasingly act through …
• We introduce KC-Bench, a controlled multi-turn benchmark for measuring…
• Its 238 tasks are manually screened from more than 1,000 generated can…

RSS 官方收录 · 可信分层展示

HalluPeer: A Taxonomy-driven Benchmark for Detecting Hallucinations in Scientific Peer Reviews

arXiv:2609.…

• 03580v1 Announce Type: new Abstract: The growing scale of academic pee…
• Existing hallucination benchmarks are not designed for peer review, wh…
• We introduce HalluPeer, a benchmark for detecting hallucinations in sc…

RSS 官方收录 · 可信分层展示

The Attention Triangle in Audio-Video Models

arXiv:2609.…

• 03586v1 Announce Type: new Abstract: Audio-video diffusion models rely…
• We study these models by probing and analyzing the ``attention triangl…
• Our analysis reveals that routing along the audio-video edge is bidire…

RSS 官方收录 · 可信分层展示

让测试更加绿色可持续

点击查看原文>

RSS 官方收录 · 可信分层展示

GPT-6 Astra 正式登场：烧了 10万块 GPU、多项跑分逼近满分，OpenAI 开启“AGI时代”

点击查看原文>

RSS 官方收录 · 可信分层展示

GPS-Bench: A Governance Policy Benchmark for Automating Policy Analysis

arXiv:2609.…

• 03553v1 Announce Type: new Abstract: Policy analysis requires more tha…
• LLM-based policy simulations model these processes at scale, but their…
• We introduce GPS-Bench, an evidence-grounded benchmark for governance …

RSS 官方收录 · 可信分层展示

Dalek: A Constructive Agent Machine

arXiv:2609.…

• 03546v1 Announce Type: new Abstract: We present Dalek, a closed machin…
• The machine is built from three primitives---actors, messages, and cha…
• Four obligations---a host boundary, a construction language, admissibl…

RSS 官方收录 · 可信分层展示

Feature Reconfiguration With Visual Prior for Medical Lesion Segmentation

arXiv:2609.03535v1 Announce Type: new Abstract: Lesion segmentation in medical images plays a critical role in clinical diagnosis and treatment planning.…

• Despite significant advances, lesion segmentation remains challenging …
• Existing encoder-decoder based methods mainly focus on enhancing featu…
• However, they lack early prior guidance and feature reconfiguration du…

RSS 官方收录 · 可信分层展示

NeoRed: A Knowledge-Logic-Alignment Multimodal Large Language Model for Neonatal Respiratory Disease Diagnosis

arXiv:2609.…

• 03527v1 Announce Type: new Abstract: Neonatal respiratory diseases are…
• Despite recent advances, existing Multimodal Large Language Models (ML…
• To address these challenges, we collect two real-world clinical datase…

RSS 官方收录 · 可信分层展示

CulturalMenuBench: Probing the Knowledge-Application Gap in Multimodal Culinary Reasoning

arXiv:2609.…

• 03526v1 Announce Type: new Abstract: Multimodal language models achiev…
• To probe this distinction, we introduce CulturalMenuBench, a benchmark…
• Evaluating 12 models exposes a substantial knowledge-application gap: …

RSS 官方收录 · 可信分层展示

一池水，困住了中国机器人

01 十秒钟2026年，一家泳池机器人创业公司CEO冯浅开始认真算一笔账。如果有传统家电公司或渠道企业愿意接手公司，多少钱可以卖？“两三倍PS，就可以谈。”几年前，他还不是这么想的。…

• 公司第一批货发往海外时，只有几百台。
• 上市前，团队在Facebook上付费招募用户填写问卷，把回收结果一条条写进产品定义。
• 按照研发团队的理解，这应该是一台更“智能”的泳池机器人：横放、竖放可以切换不同工作模式；连上App，可以控制方向、查看电量、设置定时任务。

RSS 官方收录 · 可信分层展示

神行者 8 上市 30.99 万元起，全系双电机+后轮转向，还在安全上玩出了新花样

9 月 3 日晚，FREELANDER 神行者首款车型神行者 8 正式上市，新车同时提供大五座与大六座两种布局，每种布局各有 PRO、MAX、MAX+ 三个版本。其中，大五座的上市指导价为： PRO 版 30.…

• 99 万元 MAX 版 34.
• 99 万元 MAX+ 版 38.
• 99 万元 大六座的上市指导价为： PRO 版 31.

RSS 官方收录 · 可信分层展示

速卖通Brand+加速新市场拓展，韩国多个重点类目三位数增长

继速卖通Brand+在欧盟逆势大涨97%后，Brand+在韩国也取得突破，加速跃迁品牌出海全新主场。记者获悉，今年8月，Brand+品牌商家成为速卖通韩国增长的主力军，其中迷你PC类目GMV环比6月增长超2倍，存储类目增长超150%，渔具等户外娱乐类目也在韩国实现快速增长，爆发明显。…

• 韩国站的爆发意味着在成熟的欧洲市场之外，Brand+品牌已迎来更多机会市场。
• 今年8月大促期间，速卖通Brand+把已经在欧洲验证的一站式品牌出海解决方案复制到了韩国、巴西、美国等市场，从品牌定制方案、渠道资源锁定，站…
• 渔具品牌LEYDUN是8月韩国Brand+品牌的爆单代表，早在开卖前，LEYDUN就通过“超级品牌日”锁定平台站内资源，针对韩国热门钓鱼钓法…

RSS 官方收录 · 可信分层展示

What Matters for Aggressive Decoding-Time KV Eviction? Temporal Aggregation and Ranking Preservation

arXiv:2609.…

• 03515v1 Announce Type: new Abstract: Decoding-time KV cache compressio…
• Under aggressive KV compression, we find that exponential-moving-avera…
• Value-norm and entropy variants remain highly correlated with attentio…

RSS 官方收录 · 可信分层展示

PPO-STGNN: A Proximal Policy Optimization Approach with Spatio-Temporal Graph Neural Networks for DAG Task Scheduling in Cloud-Edge-End Computing

arXiv:2609.…

• 03503v1 Announce Type: new Abstract: With the rapid development of the…
• However, cloud, edge, and end nodes are highly heterogeneous in comput…
• Traditional heuristic algorithms and conventional reinforcement-learni…

RSS 官方收录 · 可信分层展示

GrowPage: On-Demand KV Budgeting for Efficient LLM Reasoning Serving

arXiv:2609.03494v1 Announce Type: new Abstract: Long-output reasoning has made the key--value (KV) cache a critical memory bottleneck for efficient LLM serving.…

• Existing KV compression methods usually rely on a predefined per-reque…
• However, reasoning workloads exhibit substantial demand variation: dif…
• We introduce \textbf{GrowPage}, an on-demand KV budgeting framework th…

RSS 官方收录 · 可信分层展示

AutoGraphForge: Towards Automated Graph Theory Discovery

arXiv:2609.…

• 03478v1 Announce Type: new Abstract: We report on our ongoing project …
• Conjecture generation is counterexample-guided and runs in rounds: a G…
• A novelty filter of $559$ classical and folklore relations, closed und…

RSS 官方收录 · 可信分层展示

Making Every Tool Call Count: Necessary Tool-Evidence Path Rewards for Agentic Vision-Language Models

arXiv:2609.…

• 03493v1 Announce Type: new Abstract: Modern vision-language models (VL…
• To acquire this missing evidence, agentic VLMs invoke tools such as im…
• However, existing training paradigms primarily evaluate tool-use based…

RSS 官方收录 · 可信分层展示

千问办公上线首月用户数突破 3000万，企业用户占比过半

9月4日消息，阿里旗下企业级通用Agent产品千问办公上线满一个月，用户数突破3000 万，企业用户占比过半。阿里巴巴是最早布局Agent产品的公司。8月3日，阿里整合几个Agent产品推出千问办公。…

• 上线仅一个月，千问办公总用户量即超过3000万，成为AI办公赛道增长最快的产品之一。
• 同时，千问办公实现了企业市场的突破，企业用户数量占比过半，体现了阿里在企业级市场的显著优势。
• 在产品上，千问办公保持高频迭代。

RSS 官方收录 · 可信分层展示

Beyond "Made with AI": Visualizing Provenance Density to Mitigate the Transparency Penalty

arXiv:2609.03460v1 Announce Type: new Abstract: As generative AI makes polished prose cheap to produce, users can no longer rely on fluency as a proxy for truth.…

• We call this failure mode the Fluency Trap: users trust fluent halluci…
• Binary ``Made with AI'' labels respond with authorship disclosure, but…
• We propose Provenance Density, an evidence-visualization interface tha…

RSS 官方收录 · 可信分层展示

Do GUI Agents Know When Not to Act? Enabling Conflict-Aware Termination for Multimodal GUI Agents

arXiv:2609.…

• 03438v1 Announce Type: new Abstract: Graphical user interface (GUI) ag…
• A reliable agent should not only know how to act, but also when not to…
• In this work, we introduce CONFLICTGUI, a benchmark covering instructi…

RSS 官方收录 · 可信分层展示