Explore · 每周精选
发现最受关注的研究论文,追踪研究趋势,订阅感兴趣的期刊与关键词。
Xiyuan Yang, Jiaru Zou, Rui Pan +9 authors
RecursiveMAS extends recursive scaling principles from single models to multi-agent systems, enabling collaborative reasoning through iterative latent-space computations with improved efficiency and accuracy.
50 篇论文 · 按点赞排序
Meng Chu, Xuan Billy Zhang, Kevin Qinghong Lin +39 authors
World models are categorized into three capability levels and four law regimes to better understand and develop predictive environment models for AI agents across diverse domains.
Zihao Li, Jiaru Zou, Feihao Fang +6 authors
Eywa is a heterogeneous agentic framework that extends language-centric systems to scientific foundation models by integrating domain-specific models with language-based reasoning interfaces for improved performance across diverse scientific domains.
Zhengxu Yu, Yu Fu, Zhiyuan He +5 authors
OneManCompany (OMC) introduces an organizational framework for multi-agent systems that enables dynamic team assembly, governance, and improvement through portable agent identities and hierarchical decision-making processes.
Weijie Wang, Xiaoxuan He, Youping Gu +9 authors
World-R1 framework improves video generation by incorporating 3D constraints through reinforcement learning and specialized text datasets while maintaining visual quality and scalability.
V Team, Wenyi Hong, Xiaotao Gu +75 authors
GLM-5V-Turbo integrates multimodal perception as a core reasoning component for agentic tasks, demonstrating strong performance in multimodal coding and visual tool use while maintaining text-only capabilities.
Keming Wu, Zuhao Yang, Kaichen Zhang +24 authors
Visual generation models need to advance beyond appearance synthesis to incorporate structural, dynamic, and causal understanding through a five-level taxonomy spanning from atomic to world-modeling generation.
Chenkai Pan, Xinglong Xu, Yuhang Xu +6 authors
Training data can be structured as source code and evaluated through unit testing, enabling systematic debugging and improvement of language models' domain-specific capabilities.
Zaid Nasser, Mikhail Iumanov, Tianhao Li +3 authors
RADIO-ViPE is an online semantic SLAM system that provides geometry-aware open-vocabulary grounding using raw monocular RGB video without requiring calibrated inputs or depth sensors.
Yuanhao Zeng, Ao Lu, Lufei Li +3 authors
Exploratory Sampling enhances LLM generation diversity by using a lightweight distiller to predict hidden representations and bias decoding toward novel semantic patterns.
Zhiheng Liu, Weiming Ren, Xiaoke Huang +12 authors
Tuna-2 is a unified multimodal model that performs visual understanding and generation directly from pixel embeddings without pretrained vision encoders, achieving state-of-the-art performance in multimodal benchmarks.
Yiming Zhang, Jiacheng Chen, Jiaqi Tan +3 authors
ReVSI addresses flaws in current spatial intelligence evaluation by creating a validated benchmark with improved annotations and controlled frame sampling conditions.
Naibin Gu, Chenxu Yang, Qingyi Si +7 authors
Co-Evolving Policy Distillation enables unified integration of multiple expert capabilities through parallel training and bidirectional policy distillation, outperforming existing methods in multi-modal reasoning tasks.
Gal Metzer, Sagi Polaczek, Ali Mahdavi-Amiri +2 authors
Researchers developed a Semantic Progress Function to analyze and correct non-linear semantic evolution in generated media, enabling smoother transitions through semantic linearization.
Hanzhong Guo, Jie Wu, Jie Liu +6 authors
RLHF-based image editing framework introduces a chain-of-thought verification reward model that improves editing performance through fine-grained reward evaluation and reinforcement learning.
Fei Bai, Huatong Song, Shuang Sun +10 authors
ClawGym presents a scalable framework for developing Claw-style personal agents with synthetic training data, verified workspaces, and benchmark evaluation.
Gongbo Zhang, Wen Wang, Ye Tian +1 authors
Researchers developed TIDE, a framework for cross-architecture distillation of diffusion large language models that improves performance through specialized modules for distillation strength modulation, context enrichment, and cross-tokenizer objectives.
Yujun Wu, Dongxu Zhang, Xinchen Li +10 authors
Intern-Atlas presents a methodological evolution graph that captures structured relationships between research methods across AI literature, enabling automated tracking of methodological development and supporting AI-driven scientific discovery.
Yibin Luo, Shiwei Gao, Huichuan Zheng +2 authors
RoundPipe introduces a novel pipeline scheduling approach that eliminates weight binding constraints in LLM fine-tuning, enabling efficient training on consumer GPUs through dynamic stage distribution and optimized synchronization.
Qi Li, Bo Yin, Weiqi Huang +6 authors
Vision-Language-Action models present unique safety challenges due to their embodied nature, requiring unified approaches across multiple domains to address threats from data poisoning to adversarial attacks and ensuring secure deployment across various applications.
Chenxin Li, Zhengyang Tang, Huangxin Lin +8 authors
Claw-Eval-Live presents a dynamic benchmark for evaluating workflow agents that tracks evolving demands and verifies task execution through detailed logging and structured assessment methods.
Jinxiang Meng, Shaoping Huang, Fangyu Lei +17 authors
Real-world data visualization (DV) requires native environmental grounding, cross-platform evolution, and proactive intent alignment. Yet, existing benchmarks often suffer from code-sandbox confinement, single-language creation-only tasks, and assumption of perfect intent. To bridge these gaps, we introduce DV-World, a benchmark of 260 tasks designed to evaluate DV agents across real-world professional lifecycles. DV-World spans three domains: DV-Sheet for native spreadsheet manipulation including chart and dashboard creation as well as diagnostic repair; DV-Evolution for adapting and restructuring reference visual artifacts to fit new data across diverse programming paradigms and DV-Interact for proactive intent alignment with a user simulator that mimics real-world ambiguous requirements. Our hybrid evaluation framework integrates Table-value Alignment for numerical precision and MLLM-as-a-Judge with rubrics for semantic-visual assessment. Experiments reveal that state-of-the-art models achieve less than 50% overall performance, exposing critical deficits in handling the complex challenges of real-world data visualization. DV-World provides a realistic testbed to steer development toward the versatile expertise required in enterprise workflows. Our data and code are available at https://github.com/DA-Open/DV-World{this project page}.
Yanghao Zhou, Jingyu Ma, Yibo Peng +3 authors
ExoActor uses third-person video generation as a unified interface to model interaction dynamics between robots, environments, and objects, enabling task-conditioned humanoid behaviors through motion estimation and execution.
Brandon Collins, Logan Bolton, Hung Huy Nguyen +3 authors
SketchVLM is a training-free framework that enables vision-language models to generate editable SVG overlays for visual explanations, improving reasoning accuracy and annotation quality across multiple benchmarks.
Fanqing Meng, Lingxiao Du, Zijian Wu +44 authors
A benchmark for evaluating language-model agents in multi-day collaborative workflows with evolving environmental states across multiple service domains.
Jiawei Yang, Zhengyang Geng, Xuan Ju +2 authors
Fréchet Distance can be effectively optimized as a training objective when decoupling population size from batch size, leading to improved generator quality and alternative evaluation metrics.
Lei Xiong, Kun Luo, Ziyi Xia +15 authors
AutoResearchBench is a benchmark for autonomous scientific literature discovery that evaluates AI agents' ability to conduct deep and wide research tasks with high difficulty, achieving low accuracy rates even among powerful LLMs.
Shiyan Su, Ruyi Zha, Danli Shi +2 authors
DiffNR enhances neural representation optimization for CT reconstruction by integrating a single-step diffusion model with specialized conditioning layers and pseudo-reference volume generation for artifact correction.
NVIDIA, Amala Sanjay Deshmukh, Kateryna Chumachenko +214 authors
Nemotron 3 Nano Omni is a multimodal model that supports audio, text, images, and video inputs with improved accuracy and efficiency over previous versions.
Zhen Zhang, Changyi Yang, Zijie Xia +11 authors
A token-level framework called Length Value Model (LenVM) is presented that estimates remaining generation length by treating it as a value estimation problem, enabling improved length control and efficiency in autoregressive models.
北京市昌平区探索星信息技术及软件开发工作室
京ICP备2026059466号