👍 119
11/04 02:59
For a decade, progress in artificial intelligence has been measured by how much a machine can do on its own. Large foundation models now write code and run multistep tasks, and personal agents act on a person's behalf. Yet people still forget what they once knew, miss what they are not looking at, f
中文介绍 本文反思以「自主性」为尺度的AI进展,提出AI增强人类智能的新范式:AI作为「认知共生体」,与人类协作互补,而非单纯替代。作者主张重新定义进步标准,关注AI如何扩展而非取代人类认知能力,并探讨了实现该愿景的挑战与路径。
👍 107
11/05 08:00
Models for physical AI tasks, such as robotics and autonomous driving, need not reason solely in language. They can use visual plans--points, bounding boxes, and trajectories--to represent spatial goals and anticipate motion. However, autoregressive vision-language models generate these plans one co
中文介绍 VFlash提出一种基于Flash原语解码的快速视觉规划方法,用于机器人和自动驾驶等物理AI任务。传统自回归视觉语言模型逐点生成轨迹,效率低下。VFlash通过并行解码视觉原语(点、框、轨迹),大幅加速规划生成,同时保持高精度,显著提升了实时决策能力。
👍 93
11/05 08:00
Reliable robotic manipulation requires both semantic understanding and precise control in three-dimensional space. However, vision-language-action (VLA) foundations largely inherit visual representations from 2D vision-language pre-training, while spatial supervision is often introduced only during
中文介绍 SLA提出一种通过持续预训练扩展空间-语言-动作(VLA)模型的方法。现有VLA模型依赖2D视觉语言预训练特征,缺乏精确3D空间理解。SLA引入空间监督信号进行持续预训练,显著提升机器人在三维空间中的操作精度与语义理解,并在多项操作任务上刷新SOTA。
👍 81
11/01 08:00
Agent decisions depend on empirical relationships that deployment observations reveal and task descriptions alone may leave unresolved. We introduce Decoupled Predictive Decision Making (DPDM), an architecture that formulates heterogeneous agent decisions as conditional prediction with a shared pret
中文介绍 本文提出解耦预测决策(DPDM)架构,将异构智能体决策统一为条件预测问题,使用共享预训练预测器。该架构能利用部署时观测数据,解决任务描述中未明确的环境关系,提升决策泛化性与鲁棒性,为多智能体系统提供通用决策框架。
👍 66
11/03 08:00
Harness evolution from a common base exhibits Locality, Complementarity, and Nonstationarity: variants repair different failures, and their utility changes as the policy learns. Rather than seeking a universally superior harness, we introduce Harness-S to selectively rescue failures of a base execut
中文介绍 Harness-S提出从兄弟训练框架(harness)中学习,以提升智能体训练的泛化能力。针对基线harness的失败模式,该方法有选择地修复特定缺陷,而非追求普遍优越的harness。实验表明,Harness-S能有效提升策略在未见任务上的表现,并降低训练成本。
👍 51
10/31 08:00
Benchmark compression reduces the often prohibitive time and cost of model evaluation by evaluating models on a small benchmark subset that closely matches full-benchmark results. To quantify estimation error, we derive an accurate analytical approximation to the error distribution when estimating f
中文介绍 本文提出BestRand基准压缩方法,通过选择与全基准结果高度匹配的小子集,大幅降低模型评估的时间和成本。作者推导了估计误差的精确解析近似,并证明随机选择结合最优子集选择能达到稳定且低成本的压缩效果,优于现有方法。
👍 44
11/05 08:00
We present WeLM-Image, a unified image foundation model that integrates multimodal understanding into text-to-image generation, instruction-guided editing, and multi-reference synthesis. Designed for single-GPU deployment in WeChat's large-scale consumer applications, the model processes text, refer
中文介绍 WeLM-Image是一个统一图像基础模型,整合了多模态理解、文生图、指令引导编辑和多参考合成能力。该模型专为微信大规模消费级应用设计,支持单GPU部署,在多种图像生成与理解任务上展现了强大的性能与灵活性。
👍 33
11/02 08:00
Long-term memory enables large language model (LLM) agents to provide continuous personalized assistance by retaining user information across interactions. Existing systems support this process through fact extraction, organization, updating, and retrieval, but largely maintain memories as independe
中文介绍 EvoMemState提出一种演化记忆状态机制,用于个人化LLM智能体的长期记忆管理。传统方法将记忆视为独立事实,而本方法通过状态演化,动态更新和关联用户信息,实现更连续、更个性化的服务。实验显示其能有效提升多轮交互中的记忆准确性与服务质量。
👍 33
11/03 14:45
A frontier AI model is trained in an AI data center, on tens of thousands of accelerators joined by a dedicated network, and every training step ends with all of them exchanging data at once. Between racks, that network is optical, and each of its four layers, from lasers to switches, runs a fixed o
中文介绍 本文提出Halley预测控制平面,用于光互连AI数据中心网络。针对AI训练中数千加速器同步通信的挑战,Halley预测网络流量需求,动态调整光网络层配置(激光器、交换机等),减少拥塞与等待,提升训练吞吐量,并降低能耗。
👍 32
11/04 08:00
Egocentric human video offers a scalable source of manipulation priors for robot learning, yet translating raw data scale into generalizable robot policies demands rigorous data curation and training recipes. We introduce Open-HumanScale, the first fully open recipe for scaling egocentric pretrainin
中文介绍 Open-HumanScale提出首个完全开源的具身基础模型预训练配方,用于扩展人类和机器人操作经验。通过严格的数据筛选和训练策略,将大规模第一视角人类视频转化为可泛化的机器人策略。该配方在模拟和真实机器人上均取得显著性能提升,并开源以促进复现。
👍 22
11/03 08:00
Multimodal agents use web and image tools to answer questions beyond their internal knowledge. Some search tasks use images mainly to identify an initial entity, then rely on text retrieval. These tasks offer limited insight into how agents connect information across images, objects, and search step
中文介绍 VisSearch-Bench是首个评估多模态智能体长程视觉搜索能力的基准。现有任务多依赖文本检索,而本基准要求智能体跨图像、物体和搜索步骤连接信息,测试其多步视觉推理与工具使用能力。实验揭示了当前多模态智能体在长程视觉搜索中的显著不足。
👍 22
11/05 08:00
Autoregressive approaches with large language models, such as long chain-of-thought reasoning, have become the default for new AI applications because they reach higher accuracy on a wide range of tasks. In production, however, they give inconsistent answers, require parsing of outputs, and add cost
中文介绍 Drex提出一种基础决策模型,针对自回归LLM在生产环境中输出不一致、需要解析和成本高的问题。Drex采用非自回归架构,直接输出决策结果,避免长思维链的生成开销,同时保持高准确率。实验表明其在多种决策任务上性能接近或优于传统方法,且推理成本更低。
👍 20
11/04 08:00
Target-conditioned diffusion drafters predict candidate tokens in parallel, yet each new block must wait for updated target-model hidden states. This forces drafting and verification to run sequentially, limiting speedup even though each block is generated in one pass. We introduce D3D, a standalone
中文介绍 D3D提出解耦扩散草稿机制,用于异步推测解码。目标条件扩散草稿模型并行预测候选token,但需等待目标模型隐藏状态更新,导致串行瓶颈。D3D通过独立生成草稿块,实现草稿与验证异步执行,显著提升推测解码加速比,且不损失生成质量。
👍 17
11/01 22:05
LLM-based agentic systems are increasingly combined with CoT reasoning to solve tasks that require accurate reasoning, external knowledge, and multi-step deliberation. Yet, explicit CoT incurs substantial generation overhead, while implicit CoT, which replaces textual rationales with compact continu
中文介绍 TAC-CoT提出工具增强的隐式思维链方法,通过交错显式token来提升LLM智能体的推理能力。隐式CoT用紧凑连续向量替代文本推理,但缺乏工具调用能力。TAC-CoT在隐式CoT中插入工具调用token,兼顾推理效率与外部知识利用,在多步推理任务上实现更优性能与成本平衡。
👍 17
10/30 16:38
Computer-use agents increasingly tackle tasks ranging from everyday desktop activities to long-horizon professional workflows. Yet interfaces centered on screenshots and mouse-and-keyboard actions can impose a human-interaction prior, requiring repeated screen interpretation while leaving the comput
中文介绍 本文提出混合计算机使用界面,统一程序化、语义和视觉控制,用于通用智能体。传统截图和鼠标键盘交互依赖人类交互先验,效率低。新界面允许智能体直接调用程序功能、操作语义元素和视觉组件,提升长程任务执行的成功率与速度,适用于桌面和专业工作流。
👍 15
11/04 21:33
Prefill-to-Autoregressive Cross-model Transfer (PACT), a form of latent communication between language models, combines a larger model's prompt representations with a smaller model's efficient generation, offering a promising way to balance capability and inference cost. However, existing systems pr
中文介绍 本文探讨预填充到自回归跨模型迁移(PACT)的机遇与挑战。PACT将大模型的提示表示与小模型的高效生成结合,平衡能力与推理成本。作者系统分析现有系统的不足,并提出改进方向,为跨模型潜在通信技术提供理论框架和实验证据。
👍 15
11/02 08:00
Reasoning substantially improves language models' ability to solve difficult problems, but long reasoning traces make these gains expensive. Efficient reasoning should express useful analysis in fewer tokens by reducing repetition and text that adds little value to solving the problem. Progress in p
中文介绍 本文研究推理历史压缩对未来生成和智能体成本的因果效应。长推理链虽提升准确率但成本高昂。作者提出通过重写压缩冗余推理内容,发现压缩历史能降低后续生成成本,且不影响性能。因果分析表明,去除低价值推理token可显著提升效率。
👍 11
11/06 00:18
Mathematical modeling turns an open-ended real-world problem into assumptions, a model, a solution, and a defended recommendation, and its answers matter only if they are both close to a credible reference and physically feasible; existing benchmarks either score solver fragments or grade reports wi
中文介绍 本文提出Terminal-style数学建模基准,首个智能体化且可验证的数学建模评估。与现有评分片段或报告不同,该基准要求智能体产出完整建模流程,包括假设、模型、求解和论证,并验证结果与参考解的一致性及物理可行性。测试显示当前智能体在此类任务上表现不佳。
👍 11
11/02 08:00
Generative AI is billed per token, yet tokens are not equally expensive. Each token enters the KV-cache and raises the cost of those that follow, so energy grows non-linearly along a sequence whose tokens look identical. No analytical model states what a token costs, and the most accurate SOTA model
中文介绍 本文提出SIMPLE定律,解析描述LLM推理的能耗成本。现有按token计费忽略了token间的依赖(KV缓存导致非线性能耗增长)。SIMPLE定律提供精确分析模型,量化每个token的实际成本,为优化推理策略和定价机制提供理论依据,实验验证其准确性。
👍 10
11/04 08:00
Chunk-interleaved full-duplex speech language models listen and speak in fixed time chunks, and each chunk must be processed before the next one arrives. Their workload is asymmetric across chunks: listen chunks leave compute idle while awaiting audio, and speak and tool-call chunks require serial g
中文介绍 本文提出推测性监听(Speculative Listening)技术,用于块交错全双工语音语言模型。针对监听块计算空闲、说话和工具调用块串行瓶颈的问题,该方法在监听块内提前解码未来说话内容,实现计算复用。实验表明能显著降低响应延迟,提升全双工交互效率。