论文

语义视觉编码器已成为多模态理解与图像生成中语义条件的关键视觉接口。然而其最终 token 丢弃了细粒度视觉细节，导致像素重建质量较差，限制了其在图像生成与编辑等对重建敏感的任务中的应用。本工作探讨理解、生成与编辑能否在由预训练语义 ViT 构建的单一视觉表征空间中建模。我们证明，语义 ViT 的冻结 Transformer 块本身并非无法保留Semantic vision encoders have become a central visual interface for multimodal understanding and semantic conditioning in image generation. However, their final tokens discard fine-grained visual details, leading to poor pixel reconstruction and limiting their use in reconstruction-sensitive tasks such as image generation and editing. In this work, we ask whether understanding, generation, and editing can be modeled in a single visual representation space built from a pretrained semantic ViT. We show that the frozen Transformer blocks of a semantic ViT are not intrinsically unable to preserve

我们提出一种通过自适应验证任务选择来实现 LLM harness 高效优化的新方法。Harness 优化基于验证性能迭代改写 harness 代码，无需更新底层模型权重即可获得显著性能提升。然而现有方法在每次迭代中对固定的验证集进行完整评估，即便某些任务随 harness 演进区分度下降，仍产生高昂的评测成本。我们提出 Task-CoEvolve，通过应对两项挑战使验证任务与 harness 协同演进：seWe present a novel approach to efficient LLM harness optimization through adaptive validation task selection. Harness optimization iteratively rewrites the harness code based on validation performance, enabling substantial performance gains without updating the underlying model weights. Existing approaches, however, evaluate a fixed validation set in full at every iteration, incurring substantial evaluation costs even on tasks that become less discriminative as the harness evolves. We propose Task-CoEvolve, which co-evolves the validation tasks with the harness by addressing two challenges: se

检索增强 LLM 越来越多地通过抓取实时网页内容来影响日常消费推荐。这带来一种新风险：LLM 推荐系统可能会摄入被生成式引擎优化（GEO）运营者污染、用于误导其判断的网页内容。我们追问：它们在多大程度上会成为假产品的无意推广者？我们提出 FORGE（Fake Online Recommendations in Generative Environments），在本地将一组固定已抓取网页中的真实商品改写为假商品，并衡量 LLM 跨 15 个类目中 225 件真实商品推荐假商品的频率。Search-augmented LLMs increasingly mediate everyday consumer recommendations by retrieving live web content. This creates a new risk: LLM recommenders may consume web content that Generative Engine Optimization (GEO) operators have polluted to mislead them. We ask: to what extent do they become unwitting promoters of fake products? We introduce FORGE (Fake Online Recommendations in Generative Environments), which locally rewrites real products in a frozen set of retrieved web pages into fake ones and measures how often the LLM recommends the fake product, across 225 real products in 15 categor

尽管文本到 3D 生成进展迅速，在低推理成本下实现高几何保真度仍具挑战。现有文本到 3D 方法要么自回归地解码离散形状 token，要么通过扩散或流模型迭代优化全局 3D 表示。然而，自回归解码是顺序执行的且无法修正错误，而扩散与流匹配模型反复处理完整表示，使高质量生成成本日益高昂。本文提出 Block3D，一种块级扩散框架，将离散 sWhile text-to-3D generation has advanced rapidly, achieving high geometric fidelity at low inference cost remains challenging. Existing text-to-3D methods either decode discrete shape tokens autoregressively or iteratively refine global 3D representations with diffusion or flow models. However, autoregressive decoding is sequential and cannot revise errors, whereas diffusion and flow-matching models repeatedly process the full representation, making high-quality generation increasingly expensive. In this paper, we propose Block3D, a block-wise diffusion framework that partitions the discrete s

现有图像编辑框架主要沿用 text-to-image 扩散模型的训练范式。然而将该范式扩展到图像编辑时暴露出两个固有差异：一是对编辑概念粒度的关注不足，二是稀疏监督信号导致的训练低效。为应对这些问题，我们建立了一个包含超过 1,000 种细粒度编辑概念的综合分层分类体系，并构建了 ConceptEdit-12M——通过改进的合成框架生成的包含 1,200 万高质量编辑对的超大规模数据集。该 library-drExisting image editing frameworks predominantly follow the training paradigm of text-to-image diffusion models. However, extending this paradigm to image editing highlights two inherent discrepancies, specifically, the insufficient attention to edit concept granularity and the training inefficiency caused by sparse supervision signals. To address these issues, we establish a comprehensive hierarchical taxonomy featuring over 1,000 fine-grained edit concepts and build ConceptEdit-12M, a massive dataset of 12 million high-quality editing pairs via an improved synthesis framework. This library-dr

开放式真实交互允许多种有效行为：Agent 可直接回答、请求澄清、提供进度更新，或在执行前进行确认。这种灵活性打破了基于分组的 RL 的核心假设：同一分组内对比的 rollout 不再保证行为可比。因此奖励模型对交互风格的偏好可能扭曲相对优势，使优化偏向奖励偏好的行为而非情境适配的行为。我们将其形式化为奖励公平性问题，并提出 ARC（Advantage RegularizatiOpen-ended real-world interaction admits multiple valid behaviors: an agent may answer directly, ask for clarification, provide progress updates, or confirm before acting. This flexibility breaks a core assumption behind group-based RL: rollouts compared within a group are no longer guaranteed to be behaviorally comparable. As a result, reward-model preferences over interaction style can distort relative advantages and steer optimization toward reward-preferred behaviors rather than context-appropriate ones. We formalize this as a reward fairness problem and propose ARC (Advantage Regularizati

一条安全规则与一段情景日志在同一 AI Agent 上下文中争夺 token。当预算溢出时，二者以相同速率被压缩；但只有规则需要精确措辞才能保持可执行性。在 20 种生产环境 Agent 配置下，Claude Code 基于 Sonnet 4.6 的 /compact 提示在一轮压缩后保留 53% 的安全规则，五轮后仅保留 10%。我们将此现象命名为"压缩悬崖"（Compaction Cliff）。我们提出 Knowledge Triage 框架，通过对 Agent 知识库的每一行按类型分类，并为每类配置独立的保留策略来解决该问题。三种确定性 operatA safety rule and an episodic log compete for the same tokens in an AI agent's context. When the budget overflows, both are summarized at the same rate; only the rule needs exact wording to remain enforceable. On 20 production agent configurations, Claude Code's /compact prompt on Sonnet 4.6 preserves 53\% of safety rules after one compaction round and 10\% after five. We name this the Compaction Cliff. We address it with Knowledge Triage, a framework that classifies each line of an agent's knowledge base by type and routes each type through its own retention policy. Three deterministic operat

提出 TRUSTMARGIN，一种免训练、即插即用的仲裁层，利用模型自身的似然对两个候选进行打分，在不微调、无需外部评判或额外生成的情况下，在直接回答与 RAG 之间进行选择。TRUSTMARGIN is proposed, a training-free, plug-and-play arbitration layer that scores the two existing candidates with the model's own likelihoods and selects between Direct and RAG without fine-tuning, external judges, or additional generation.

真实工业场景中的用户表示学习通常通过增加用户数量、行为序列长度和模型规模来扩展。然而现有方法面临两个挑战：(i) 十亿级容量下原始数据扩展的瓶颈，因为随着更大规模原始文本用户行为输入的增加，性能增益呈现递减趋势，这可以通过 tokenization 缓解；(ii) 缺乏对 tokenization 配置应如何随数据规模扩展的定量分析。本报告中，我们提出 User Behavioral Densing Law 来刻画定量关系User representation learning in real-world industrial scenarios is commonly scaled by increasing user amount, behavioral sequence length and model size. However, existing methods face two challenges: (i) Bottleneck for raw data scaling at billion-scale capacity, as performance exhibit diminishing performance gains with larger-scale raw text user behavioral input, which can be mitigated by tokenization. (ii) Lack of quantitative analysis of how tokenization configurations should scale with data size. In this report, we propose User Behavioral Densing Law for characterizing the quantitative rela

大语言模型日益被期望执行复杂工作流，其成功依赖于维持相互依赖的约束并产出满足严格端到端验证的产物。然而成功的执行经验通常在单次运行后就丢失了，迫使后续模型从头重新发现策略和失败模式。我们研究能否通过 EvoMap 将此类经验外部化并复用，其中验证器确认的执行轨迹被整合为结构化的 Gene。为评估该设定，我们引入长工作流基准Large language models are increasingly expected to execute complex workflows whose success depends on maintaining interdependent constraints and producing artifacts that satisfy strict end-to-end verification. Yet successful execution experience is typically lost after a single run, forcing subsequent models to rediscover strategies and failure modes from scratch. We study whether such experience can instead be externalized and reused through EvoMap, where verifier-confirmed execution trajectories are consolidated into structured Gene. To evaluate this setting, we introduce the Long-Workflow B

自主研究系统执行长研究工作流的能力日益增强，但仅靠自动化并不能确保所得流程保持科学严谨性。我们提出 AutoResearch，一个两阶段系统，将 Idea Generation 与 Idea Execution 相连，以同时解决研究想法如何形成以及如何通过实验可靠建立的问题。在 Idea Generation 中，AutoResearch 持续整合新出现的研究信号与累积的领域知识，识别可迁移的机理洞察，并采用多模型生成与跨模型Autonomous research systems are increasingly capable of executing long research workflows, yet automation alone does not ensure that the resulting process remains scientifically grounded. We introduce AutoResearch, a two-stage system that connects Idea Generation with Idea Execution to address both how research ideas are formed and how they are reliably established through experimentation. In Idea Generation, AutoResearch continuously integrates emerging research signals with accumulated domain knowledge, identifies transferable mechanistic insights, and uses multi-model generation and cross-r

深度人脸识别 (FR) 模型已达到近乎饱和的精度，但仍缺乏透明度：从业者无法追问某个相似度分数究竟依赖了哪些语义属性。EXPL-FR 在 FR 模型自身的嵌入空间内给出答案。一个轻量适配器将 vision-language model (VLM) 的图像编码器与冻结的 FR 空间对齐，仅基于人脸图像训练，从不基于文本。由于 VLM 的编码器共享同一空间，同一适配器同样适用于文本编码器，从而无需额外成本即可将 22 个类别中的 978 条属性提示（也可扩展）转化为 FR 空间锚点。我们并未假设这种迁移Deep face recognition (FR) models reach near-saturated accuracy but remain opaque: a practitioner cannot ask which semantic attributes a similarity score relied upon. EXPL-FR answers this inside the FR model's own embedding space. A lightweight adapter aligns a vision-language model's (VLM) image encoder with the frozen FR space, trained on face images alone and never on text. Because the VLM's encoders share one space, the same adapter applies to the text encoder, turning 978 attribute prompts in 22 categories, also extendable, into FR-space anchors at no extra cost. We do not assume this tra

本工作开发了一个 SOTA 的 paged attention kernel，完全基于领域特定即时编译语言 Triton 构建，在 NVIDIA 与 AMD GPU 上均达到 SOTA 性能。This work develops a state-of-the-art paged attention kernel that builds exclusively on the domain-specific just-in-time compiled language Triton to achieve state-of-the-art performance on both NVIDIA and AMD GPUs.

本工作提出一种新颖的混合精度 PTQ 策略，直接最小化整个模型的全局误差传播，而非孤立地处理逐层误差，并开发了一种新颖的联合优化方法，在统一搜索空间中同时学习结构化剪枝决策与混合精度量化策略。This work proposes a novel mixed-precision PTQ strategy that directly minimizes global error propagation across the entire model, rather than isolating layer-wise errors, and develops a novel joint optimization approach that simultaneously learns structural pruning decisions and mixed-precision quantization policies within a unified search space.

本工作提出 QBugLM，一个多 Agent 框架，可自动化量子软件调试流水线，覆盖基于分类法的缺陷注入、基于 LLM 的检测与修复，直至基于仿真的验证，框架无关地支持 OpenQASM 3.0 程序。This work proposes QBugLM, a multi-agent framework that automates the quantum software debugging pipeline, from taxonomy-driven bug injection to LLM-based detection and repair, and finally to simulation-based validation, for framework-agnostic OpenQASM 3.0 programs.

PLENA 是一个软硬件协同设计的系统，采用三条核心优化路径，具备新颖的扁平化 systolic-array 架构以及支持非对称量化方案的高效计算与存储单元（路径 2）。PLENA is a hardware-software codesigned system that applies three core optimization pathways that features a novel flattened systolic-array architecture and efficient compute and memory units that support an asymmetric quantization scheme (Pathway 2).

本工作将 embedding 缩放作为正交于稀疏度缩放的强有力维度加以探索，并推出 LongCat-Flash-Lite，一个从零训练的 68.5B 参数、约 30 亿激活参数的模型，不仅超越参数等量级的 MoE 基线，还对同规模现有模型展现出卓越竞争力。This work explores embedding scaling as a potent, orthogonal dimension for scaling sparsity and introduces LongCat-Flash-Lite, a 68.5B parameter model with ~3B activated trained from scratch that not only surpasses parameter-equivalent MoE baselines but also exhibits exceptional competitiveness against existing models of comparable scale.

ViT-5 的设计与当代基础模型实践保持一致，可作为对 vanilla ViT 的直接替换升级方案，适用于 2020 年代中期的视觉骨干网络，并为生成建模提供更强大的骨干。With a design aligned with contemporary foundation-model practices, ViT-5 offers a simple drop-in upgrade over vanilla ViT for mid-2020s vision backbones and serves as a stronger backbone for generative modeling.

GLM-5 在真实编码任务中展现出前所未有的能力，在端到端软件工程挑战的处理上超越既有基线，并提出了新颖的异步 Agent RL 算法，进一步提升了 RL 质量。GLM-5 demonstrates unprecedented capability in real-world coding tasks, surpassing previous baselines in handling end-to-end software engineering challenges and proposing novel asynchronous agent RL algorithms that further improve RL quality.

本工作借鉴线性模型的 state space model（SSM）视角，提出三项核心方法改进并组合形成更具表达力的递推结构：源自 SSM 离散化的递推式、用于更丰富状态追踪的复数值状态更新规则，以及在不增加 decode 延迟前提下提升模型性能的多输入多输出（MIMO）建模。This work introduces three core methodological improvements inspired by the state space model (SSM) viewpoint of linear models that combine to form a more expressive recurrence derived from SSM discretization, a complex-valued state update rule that enables richer state tracking, and a multi-input, multi-output (MIMO) formulation for better model performance without increasing decode latency.

Nemotron 3 Super 是 Nemotron 3 系列中首个采用 NVFP4 进行预训练的模型，借助 LatentMoE（一种同时优化精度 per FLOP 与精度 per parameter 的新型 Mixture-of-Experts 架构），并集成 MTP 层以通过原生 speculative decoding 加速推理。Nemotron 3 Super is the first model in the Nemotron 3 family to be pre-trained in NVFP4, leverage LatentMoE, a new Mixture-of-Experts architecture that optimizes for both accuracy per FLOP and accuracy per parameter, and include MTP layers for inference acceleration through native speculative decoding.

本教程综合了推动这些汇聚方向的算法、系统与设计原则，为数据科学与数据挖掘研究者提供统一视角，涵盖将 LLM、图数据管理、图挖掘、图 ML 与 agentic 计算融合到下一代 graph-native AI 系统中。This tutorial synthesizes the algorithms, systems, and design principles driving these converging directions, offering data science and data mining researchers a unified perspective on integrating LLMs, graph data management, graph mining, graph ML, and agentic computation into next-generation graph-native AI systems.

本文提出 LoRAFusion，一种面向 LLM 的高效 LoRA 微调系统，可消除不必要的内存访问，在不付出重算或同步代价的前提下保持 compute-bound GEMM 的性能，并引入面向多任务微调的自适应批处理算法。LoRAFusion is introduced, an efficient LoRA fine-tuning system for LLMs that eliminates unnecessary memory accesses and preserves the performance of compute-bound GEMMs without incurring the cost of recomputation or synchronization and introduces an adaptive batching algorithm for multi-job fine-tuning.

Tangram 是一种 serving 框架，将先前系统动态处理的内容静态解析，可作为现有 non-uniform 压缩方法的即插即用底座，在匹配其精度的同时，端到端吞吐量较 full-KV 基线最高提升 2.6×。Tangram is a serving framework that statically resolves what prior systems handle dynamically, and serves as a drop-in substrate for existing non-uniform compression methods, matching their accuracy while improving end-to-end throughput by up to $2.6\times over the full-KV baseline.

本文认为，AI agent——即以大语言模型作为主要推理引擎、动态生成与丢弃代码作为工具性资源的系统——的出现构成了对"软件"本身的根本性重构，而非渐进式的工具改进。This paper argues that the emergence of AI agents -- systems where large language models serve as the primary reasoning engine, dynamically generating and discarding code as an instrumental resource -- constitutes a fundamental restructuring of what software is, not an incremental tool improvement.

本文提出 Memanto，一种面向 agentic 人工智能的通用记忆层，挑战了"必须依赖知识图谱复杂度才能实现高保真 agent 记忆"的普遍假设，并取得 SOTA 准确率。Memanto is introduced, a universal memory layer for agentic artificial intelligence that challenges the prevailing assumption that knowledge graph complexity is necessary to achieve high fidelity agent memory and achieves state of the art accuracy scores.

本文提出 MatryoshkaLoRA，一种受 Matryoshka 启发、面向 LoRA 的通用训练框架，通过在已有 LoRA adapter 之间插入一个固定的、经精心设计的对角矩阵来按比例缩放其子秩，从而学习到准确的层次化低秩表示。MatryoshkaLoRA is proposed, a general, Matryoshka-inspired training framework for LoRA that learns accurate hierarchical low-rank representations by inserting a fixed, carefully crafted diagonal matrix between the existing LoRA adapters to scale their sub-ranks accordingly.

本文为 LLM serving 中的 SD（投机解码）提出了一种简单且可解释的 latency 模型，能够准确刻画实际观测到的 latency，解释为何加速比常随服务器负载上升而下降，并系统刻画了 draft length、acceptance rate 以及 verifier 与 drafter 规模在不同 serving 条件下对 latency 的影响。A simple and interpretable latency model for SD in LLM serving is developed that accurately describes observed latency, explains why speedups often diminish as server load increases, and characterizes how draft length, acceptance rate, and verifier-drafter size shape latency across serving conditions are characterized.

Tutti 是一种高效的 SSD-backed KV caching 方案，将 CPU 从 HBM 与 SSD 之间的关键数据与 I/O 控制路径中彻底移除，在提供近乎无限容量的同时，实现了与 DRAM-backed LMCache 几乎相当的 inference 性能。Tutti is an efficient SSD-backed KV caching solution that eliminates CPU intervention from the critical data and I/O control paths between HBM and SSDs, and achieves nearly the same inference performance as DRAM-backed LMCache, while providing almost infinite capacity.

这些结果表明，Kueue、DAS 与 GAIE 等互补组件构成了一个高性能的协同平台，证明了 Kubernetes 能够作为承载高要求 GenAI 工作负载的统一底座。These findings illustrate that these complementary components (Kueue, DAS, and GAIE) form a cohesive, high-performance platform, proving Kubernetes' capability to serve as a unified foundation for demanding GenAI workloads.

本文提出了一种简单的 flow-control 框架，通过控制 prompt 加入 LLM 活跃集合的速率，实现更高的 token 与 request 吞吐量、更低的平均与尾部 latency，以及更稳定的 KV cache 利用率。A simple flow-control framework is proposed that controls the rate at which prompts join the active set in large language models and achieves higher token and request throughput, lower average and tail latency, and more stable KV cache utilization.

本文提出 CATS，一种 self-speculative decoding 框架，在内存受限设备上结合 memory budget 与 parameter offloading pattern 进行级联式 verify 与 correction，在设备峰值显存与单独运行 target model 相当的前提下，最大化 token acceptance rate 与端到端加速比。CATS, a self-speculative decoding framework that conducts cascaded verification and correction based on the memory budget and parameter offloading patterns on memory-limited devices, is proposed, which maximizes token acceptance rate and end-to-end speedup while keeping the peak memory footprint on the device equal to that of the target model alone.

本文在 Argonne Leadership Computing Facility 的 Polaris 超级计算机上对分布式向量数据库性能进行了实证研究，选取 Qdrant 评估在最多 32 个 worker 下的插入、索引构建与 query latency。This work presents an empirical study of distributed vector database performance on the Polaris supercomputer in the Argonne Leadership Computing Facility, and selects Qdrant to evaluate insertion, index construction, and query latency with up to 32 workers.

本文提出 DualPath，一种 inference 系统，通过引入 dual-path KV-Cache loading 打破瓶颈，并实现一条新的 storage-to-decode 路径：KV-Cache 先加载到 decode engine，再通过 compute network 上的 RDMA 高效转发至 prefill engine。DualPath is presented, an inference system that breaks this bottleneck by introducing dual-path KV-Cache loading and enables a novel storage-to-decode path, in which the KV-Cache is loaded into decoding engines and then efficiently transferred to prefill engines via RDMA over the compute network.

本文通过大规模超参数搜索，系统地重新评估了 Vanilla LoRA 以及九个代表性 LoRA 变体，发现不同 LoRA 方法偏好的学习率区间各异，并将最优学习率区间的差异归因于最大 Hessian 特征值的变化，与经典学习理论相吻合。This work systematically re-evaluate nine representative LoRA variants alongside vanilla LoRA through extensive hyperparameter searches, finding that different LoRA methods favor distinct learning rate ranges and attributes the differing optimal learning rate ranges to variations in the largest Hessian eigenvalue, aligning with classical learning theories.

本文提出 TTKV，一种 KV cache 管理框架，将人类记忆系统映射到具备异构容量与精度的 KV cache 上，在 128K 上下文任务上将跨层流量降低 5.94 倍。TTKV is proposed, a KV cache management framework that maps the human memory system onto the KV cache with heterogeneous capacity and precision, and reduces cross-tier traffic by 5.94x on 128K-context tasks.