👍 135
11/11 08:00
System One models promise fast, general-purpose decision-making without lengthy token generation. However, understanding their capability spectrum and decision reliability requires more than aggregate performance on current benchmarks. Existing evaluations often provide limited insight into the brea
中文介绍 针对 System One 模型的快速通用决策能力,SO-Arena 提出评估其能力谱与决策可靠性的基准。现有评测仅给聚合分数,难以揭示模型在不同推理强度下的边界表现,该工作通过更细粒度的指标刻画 System One 模型的决策能力范围。
👍 47
10/29 08:00
Mathematical supervision for language models is typically flattened into isolated problem-solution pairs, discarding the curricular organization and evidential structure of the original sources. We introduce TOME-Math (Textbook-Oriented Mathematical Evidence), a large-scale corpus that converts 2,11
中文介绍 数学推理训练常将教材压成孤立题解对,丢失课程结构与证据链。本文提出 TOME-Math(Textbook-Oriented Mathematical Evidence),从 2,112 本结构化学术教材构建大规模语料,保留章节层次与证明逻辑,用于为语言模型提供更具上下文的数学监督信号。
👍 46
11/12 08:00
Video generation models produce visually realistic videos but still struggle with physical dynamics and causal consistency. Learning from generated videos requires fine-grained physical supervision, yet reliable quality judgments, correction pairs, and controlled variations are difficult to obtain.
中文介绍 PhysLoop 利用 LLM 的物理常识为视频生成模型提供细粒度物理监督。通过自动判断生成视频的质量、构造纠正样本对及可控变体,形成自改进闭环,改善生成内容在物理动力学与因果一致性上的不足,减少人工标注需求。
👍 40
11/11 08:00
In this report, we introduce the Reciprocal World Model (RWM) and present Duo v1, a full-duplex world model that continuously perceives audiovisual inputs while responding to users in real time. The RWM treats intelligence and physical reality as equally fundamental to what constitutes a world. Exis
中文介绍 Duo 提出互惠世界模型(RWM)与首个全双工世界模型 Duo v1,可持续感知视听输入并实时响应用户。RWM 认为智能与物理现实同等重要,将两者结合建模,支持边听边说的连续交互,突破现有单模态、半双工语音系统的局限。
👍 36
11/10 08:00
Low-bit quantization reduces the inference cost of LLM agents but can compromise task success. Examining failed rollouts, we find that quantized agents overlook task requirements or environmental conditions, producing environment transitions that diverge from successful full-precision references. Co
中文介绍 SteerDistill 针对低比特量化 LLM Agent 任务成功率下降的问题,分析失败轨迹发现量化 Agent 忽略任务要求或环境条件,导致状态转移偏离全精度参考。方法通过闭环教师指导(Closed-Loop Teacher Steering)校正偏差,实现量化部署下的稳定推理。
👍 35
11/11 08:00
Recent work suggests that pretrained image generators encode real-world understanding beyond image synthesis. Vision Banana exploits this by framing vision tasks as conditional image editing. However, its data, weights, code, and implementation details are not public. We present Open Vision Banana (
中文介绍 Vision Banana 将视觉任务重构为条件图像编辑,利用预训练图像生成器中的世界理解能力,但其数据与权重未公开。Open Vision Banana 提供完整开放配方,包括数据、权重与代码,复现并验证该生成式视觉感知范式的可行性。
👍 33
11/12 08:00
Decision models answer a question with a typed output, such as a choice among options with a probability for each, so one model can choose an agent's actions, evaluate actions already taken, or score every action to teach a policy with reinforcement learning. Prior work has studied them mainly as ac
中文介绍 决策模型在长程 Agent 中可作执行者、评判者或教师,但现有研究主要限于其作为 Actor 的角色。本文系统比较同一模型在三种决策模式下的表现,分析其选择动作、评估已执行动作、以及通过强化学习打分训练策略的能力差异。
👍 29
11/04 08:00
Hyper-Connections (HC) architectures are increasingly adopted in frontier language models, yet their components remain insufficiently understood and introduce non-negligible training overhead. In this work, we systematically study the HC architecture to understand its components and identify simpler
中文介绍 Hyper-Connections(HC)架构已被前沿语言模型采用,但组件理解不足且带来可观训练开销。本文系统研究 HC 架构,识别并简化非必要组件,在保持性能的同时提升训练稳定性与可扩展性,降低推理与训练成本。
👍 22
11/12 08:00
Decoder-only Large View Synthesis Models (LVSMs) with KV-cache have recently achieved state-of-the-art quality by regressing novel views from neural networks without reconstructing 3D geometry, while Test-Time Training (TTT) layers save computation but sacrifice quality. We propose Test-Time Baking
中文介绍 仅解码器视角合成模型(LVSM)借助 KV-cache 在不重建 3D 几何的情况下回归新视角,画质达 SOTA;但 Test-Time Training(TTT)层虽省计算却损失质量。TTB 提出 Test-time Baking,兼顾计算效率与合成质量。
👍 22
11/07 08:00
Web-facing AI agents repeatedly access webpages, documents, repositories, and tool responses, yet reading interfaces often expose entire objects when only sparse evidence is needed. This over-reading increases token and latency costs and can dilute relevant evidence. We present SparseRead, a reusabl
中文介绍 面向 AI Agent 的网页与文档读取常暴露完整内容,而实际只需稀疏证据,造成 token 浪费与关键信息稀释。SparseRead 提出可复用的稀疏读取接口,按证据相关性裁剪内容,显著降低 token 消耗与延迟,提升 Agent 推理质量。
👍 18
11/05 08:00
Self-improving embodied agents are increasingly being developed and compared on robot simulation benchmarks. Several recent works expose the benchmark's success criteria to the learning agent, a practice we call Criteria Exposure (CE). These criteria, originally written to verify policies trained by
中文介绍 自改进具身 Agent 的基准评测中,部分工作向学习体暴露成功判据(Criteria Exposure, CE),这些判据原用于验证策略而非指导学习。本文重新审视该做法,分析暴露判据对 Agent 能力提升的实际贡献与对评测可信度的影响。
👍 16
11/12 08:00
Multimodal coding agents decouple 4D world construction from visual generation, repurposing video models from world generators toward generative renderers. This division of labor raises a fundamental question: if code builds the world, can video models just render it? A reliable generative renderer
中文介绍 多模态编码 Agent 将 4D 世界构建与视觉生成解耦,使视频模型从世界生成器转变为生成式渲染器。WorldDraft 系统评估该分工模式:若代码负责构建世界状态,视频模型能否可靠执行渲染任务?研究提出评估框架与关键发现。
👍 14
11/12 08:00
How multimodal language models should encode visual input remains up for debate. Most models use a pretrained vision encoder, while some recent recipes train the encoder from scratch with the language model or remove it entirely. Vision encoders bundle together three design choices: a visual prior f
中文介绍 多模态模型如何编码视觉输入尚无共识:主流方案用预训练视觉编码器,部分工作从头训练或直接移除。Unbundling 将视觉编码器拆解为视觉先验、特征压缩与对齐三个独立设计选择,分别评估各组件的必要性。
👍 14
11/12 08:00
Second-order optimizers (e.g., Shampoo, SOAP) have emerged as compelling alternatives to Adam/AdamW and promise faster per-step convergence than AdamW in transformer training. However, they still remain costly in practice, and even SOAP still pays substantial optimizer-state memory and eigendecompos
中文介绍 二阶优化器(如 Shampoo、SOAP)在 Transformer 训练中每步收敛快于 AdamW,但优化器状态内存与特征分解开销仍高。Factored One-sided Adam-Moment 提出分块单边近似,降低 SOAP 的实际计算与存储成本,提升可扩展性。
👍 12
11/09 08:00
Agentic workloads are a major source of LLM inference demand. Existing prefill--decode (PD) disaggregated serving systems offer limited flexibility for runtime adaptation of compute allocation and KV cache management, and rely on coarse signals that only partly capture the diverse and dynamic agent
中文介绍 Agentic 工作负载是 LLM 推理的主要需求来源,但现有 prefill-decode 分离式服务系统在运行时计算分配与 KV cache 管理上灵活性不足。Evolving Infra 提出自适应 LLM 服务架构,依据 Agent 多样化动态需求实时调整资源调度。
👍 7
11/07 08:00
Self-evolving agents improve their own skills, memory, and harness code from experience without retraining the base LLM, and group-evolving agents go further by sharing these lessons across agents. Yet most such systems overlook that sharing across users and heterogeneous agents requires deciding wh
中文介绍 自进化 Agent 从经验中改进自身技能与记忆,群体进化 Agent 进一步跨体共享。本文指出跨用户与异构 Agent 共享需决定「记住什么」,提出多粒度记忆抽取方法,按记忆粒度层级管理共享,提升多用户多 Agent 交互中的知识复用效率。
👍 6
11/12 08:00
While search agents demonstrate impressive capabilities in multi-step question answering, their robustness to poor-quality evidence remains under-explored. Such evidence is rare in benchmark corpora, and resulting failures are hard to reproduce. We propose DeepStress, a dynamic stress-testing framew
中文介绍 搜索 Agent 在多步问答中表现出色,但对低质量证据的鲁棒性仍未被充分检验。DeepStress 提出动态压力测试框架,向 Agent 注入罕见且难以复现的低质证据,评估其在干扰条件下的推理稳定性与抗噪能力。
👍 6
11/04 08:00
Subliminal learning is a phenomenon in large language models (LLMs) where a teacher passes a hidden preference to a student through training data unrelated to that preference. We study whether subliminal learning also occurs in vision-language models (VLMs). We finetune a teacher VLM to prefer a spe
中文介绍 潜学习(Subliminal Learning)指教师通过与学生偏好无关的训练数据传递隐藏偏好,该现象已在 LLM 中证实。本文将研究扩展至视觉语言模型(VLM),通过微调教师 VLM 偏好特定视觉属性,验证潜学习在 VLM 中是否同样成立。
👍 5
11/10 08:00
Local decoding usually fixes one weight precision for the whole generation. We study a different allocation: store 8-bit weights, decode with their top 4 bits, and redo a step at 8 bits only when the low-precision output distribution looks uncertain (entropy above a threshold or a small top-two marg
中文介绍 本地推理通常固定单一权重精度。本文提出按需升级策略:以 8-bit 存储权重、4-bit 解码,仅在低精度输出分布不确定(熵超阈值或 top-2 边际概率接近)时,对该步重新以 8-bit 计算,平衡速度与精度。
👍 4
11/12 08:00
Reference-frame stabilization registers every frame of a hovering-drone video onto one reference frame, keeping object positions comparable. The registration pipelines it relies on are selected on feature-matching benchmarks in which only the camera moves. Over a busy intersection, vehicles moving t
中文介绍 基于参考帧的无人机视频稳定将悬停画面各帧对齐至单一参考帧,保证物体位置可比。现有配准流水线依据仅相机运动的特征匹配基准选择,在繁忙路口等动态城市场景下面临车辆与行人移动带来的挑战,本文提出相应基准。