👍 64
07/04 08:00
Optimizer selection for large-scale model training has become a system-level design decision constrained jointly by compute, memory, tuning budget, and task diversity, yet the landscape of over one hundred methods remains fragmented. We therefore present OmniOpt, a unified survey and benchmark cookb
中文介绍 提出OmniOpt,统一分类与几何视角审视百余种优化器,结合计算、内存、调优预算和任务多样性进行系统级基准测试,为大规模模型训练提供实用指南。
👍 62
07/05 08:00
Recent advances in multimodal foundation models and agent systems have driven GUI agents from single-platform task execution toward cross-platform interaction. However, building multi-platform GUI agents remains challenging. On one hand, high-quality and executable cross-platform interaction traject
中文介绍 提出UI-MOPD,多平台在策略蒸馏框架,通过跨平台交互轨迹的在线策略学习持续更新GUI Agent,解决多平台泛化与数据稀缺问题。
👍 47
07/05 08:00
Research dissemination, turning a paper into a poster, a talk video, and a blog post, is still a manual last mile. Prior automation treats each artifact in isolation that each re-extract the paper from scratch, usually ship one-way renders the author cannot reopen in PowerPoint or Word, and gates qu
中文介绍 ResearchStudio-Reel自动化研究传播最后环节,将论文一键转化为海报、视频和博客,支持可编辑输出(PowerPoint/Word),打破各制品孤立生成与单向渲染局限。
👍 46
07/06 08:00
3D reconstruction and generation are commonly tackled by separate paradigms: pixel-based regression for reconstruction, and latent diffusion for generation. Recent works attempt to unify them in latent space, but with notable drawbacks: the diffusion objective is defined on latent features rather th
中文介绍 PixWorld在像素空间统一3D场景生成与重建,将扩散目标定义于RGB像素而非隐空间,避免特征域失真,同时支持重建与生成的协同训练。
👍 31
07/02 08:00
Evaluating embodied robot foundation models remains a critical bottleneck; unlike large language models efficiently assessed via digital benchmarks, robotic policies require slow, costly real-world rollouts limited by hardware and human supervision, which has driven interest in world models as surro
中文介绍 提出GigaWorld-1路线图,构建世界模型作为机器人策略评估的替代性数字环境,降低对昂贵真实场景回滚的依赖,加速具身基础模型评测。
👍 30
07/06 08:00
Dense spatial perception is essential for physical intelligence, where visual systems are expected to recover structured, metric, and actionable representations from pixel observations. Modern visual foundation models tend to prioritize semantic invariance, often at the expense of detailed spatial u
中文介绍 针对物理智能中密集空间感知需求,研究视觉预训练如何平衡语义不变性与几何结构恢复,提出面向度量、可操作的密集表征的训练方案。
👍 24
07/05 08:00
We present Wan-Streamer v0.2, a latency-preserving upgrade of the native-streaming, end-to-end audio-visual interaction model. v0.2 keeps the v0.1 modeling formulation, but raises the interactive output stream from 192x336 to 640x368 while preserving approximately 200 ms model-side signal-to-signal
中文介绍 Wan-Streamer v0.2保持原生流式端到端音视频交互延迟约200ms,将输出分辨率从192×336提升至640×368,实现低延迟高画质升级。
👍 18
07/01 08:00
Scientific literature search often requires more than retrieving papers from a single query: users' intents are underspecified, preference-dependent, and evolve through interaction. Existing search agents typically rely on fixed pipelines or implicit language-only reasoning, making their search stra
中文介绍 提出多轮科学文献搜索Agent,通过工作流归纳根据用户隐性意图动态调整检索策略,超越固定流水线与纯语言推理,提升搜索透明性与适应性。
👍 16
07/06 08:00
Key-value (KV) cache growth is a major bottleneck in autoregressive decoding, as memory and bandwidth scale linearly with context length. Existing KV eviction methods often rely on static heuristics or proxy scores, which poorly track future token utility and cause brittle eviction as relevance shif
中文介绍 KVpop采用预测性在线剪枝压缩Key-Value缓存,根据未来token效用动态驱逐低值条目,克服静态启发式在长上下文中的固定失效问题。
👍 15
07/06 08:00
Multi-vector vision-language retrieval preserves fine-grained visual evidence through maximum-similarity late interaction, but dense image-side tokens make storage and scoring expensive. Existing token compression methods reduce this cost, yet they can remove or collapse object- and region-level evi
中文介绍 提出目标证据保留的token合并方法,在视觉-语言检索中保留对象区域证据,减少密集图像token存储与评分成本,同时保持多向量晚期交互性能。
👍 14
07/05 08:00
Diffusion large language models (dLLMs) generate text by iteratively denoising a masked sequence, offering a parallel alternative to autoregressive models, but eliciting strong reasoning through post-training remains difficult: supervised fine-tuning is off-policy and suffers from exposure bias, whi
中文介绍 dOPSD将在线自我蒸馏用于扩散语言模型后训练,缓解监督微调的分布偏移,通过动态掩码采样优化推理质量,提升文本生成与逻辑推理能力。
👍 13
07/06 08:00
We introduce the first multiplayer world model for highly dynamic environments governed by complex physical interactions. Whereas single-player world models treat the other agents as part of the environment, ours conditions on the action streams of multiple agents, learning to attribute changes in t
中文介绍 首个多玩家交互式世界模型,以表示自编码器处理多智能体动作流,在复杂物理动态环境中将玩家间的相互影响与场景变化解耦建模。
👍 11
07/06 08:00
Pretraining scaling laws reveal that model capability improves predictably with data and compute. But learning from real world environments after deployment remains far less understood. Analyzing roughly 38,000 hours of agent interaction with the environment across 134 real world tasks, we find, to
中文介绍 EdgeBench基于约38000小时真实环境交互数据,揭示自监督适应中模型能力随数据与计算增长的比例定律,为部署后学习提供实证基础。
👍 11
06/30 08:00
The fast growth of open-source AI infrastructure, from model serving engines and agent platforms to the Model Context Protocol (MCP) ecosystem and the language models themselves, has outpaced the security tooling available to defend it. We present AI-Infra-Guard, an open-source framework that organi
中文介绍 AI-Infra-Guard开源框架,覆盖模型服务引擎、Agent平台、MCP协议和语言模型多层安全,提供统一红队测试工具集以应对快速增长的AI基础设施威胁。
👍 10
07/01 08:00
LLMs are increasingly used to brainstorm research ideas, but existing evaluations mostly judge individual ideas by novelty, feasibility, or expert preference. We instead ask: how far are current LLM-generated ideas from human researchers? To characterize this gap, we build a large-scale evaluation f
中文介绍 大规模评估框架衡量LLM生成研究想法与人类研究者的差距,从多样性、可行性等维度量化,揭示当前模型在创意新颖性上的不足。
👍 9
07/06 08:00
Scaling pre-training, post-training, and test-time compute have become the central paradigms for improving the capabilities of LLMs. In this work, we identify verification, the ability to determine the correctness of a solution, as a new scaling axis. To unlock this and demonstrate its effectiveness
中文介绍 提出LLM-as-a-Verifier通用验证框架,将“验证”作为独立缩放轴,通过训练验证器判断解决方案正确性,与预训练、后训练及测试时计算互补提升模型能力。
👍 9
07/01 08:00
As grounded QA systems are increasingly deployed in AI assistants, accurately attributing generated answers to evidence is critical for user trust and model safety. While unimodal attributions have been explored in depth, the multimodal setting remains relatively under-researched. As a result, we in
中文介绍 MultAttnAttrib无训练多模态归因方法,在长文档问答中为生成答案提供基于注意力证据的跨模态溯源,无需额外训练即可提升可信度。
👍 7
07/06 08:00
Audio intelligence involves understanding, reasoning about, and generating both audio and speech. In this work, we introduce Nemotron-Labs-Audex-30B-A3B (Audex), a unified audio-text LLM built on Nemotron-Cascade-2-30B-A3B, a strong text-only MoE LLM. Audex adopts a simple unified design with a sing
中文介绍 Audex统一音频-文本LLM,基于Nemotron-Cascade-2 MoE构建,单框架支持理解、推理与生成音频/语音,且不退化文本能力。
👍 6
06/30 08:00
Large language models increasingly operate over long contexts, where the KV cache becomes a dominant memory bottleneck: its size grows linearly with sequence length and must be retained throughout decoding, making full GPU caching prohibitively expensive without compression. Existing KV cache compre
中文介绍 SeKV通过分辨率自适应KV缓存与层次化语义记忆压缩,在长上下文LLM推理中降低GPU内存占用,保持解码质量。
👍 6
07/03 08:00
Long-horizon behavior prediction aims to infer a user's next action based on a lengthy historical sequence, playing a crucial role in artificial intelligence field. The rise of large language models (LLMs) offers a promising direction for sequential behavior prediction, yet LLMs struggle with latent
中文介绍 PraMem为长时空行为预测构建实践经验记忆模块,增强LLM对隐性偏好的挖掘能力,提升从历史序列推断下一步动作的准确性。