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。
50 篇论文 · 按点赞排序
Jinyang Wu, Shuo Yang, Zhengxi Lu +8 authors
SEED improves long-horizon agent training by extracting hindsight skills from completed trajectories and distilling them into dense token-level supervision alongside outcome-based reinforcement learning.
Ruiyan Gong, Yingnan Guo, Junjun Hu +43 authors
ABot-N1 improves visual language navigation by separating reasoning from control through a slow-fast architecture that uses explicit chain-of-thought reasoning and pixel goals to guide continuous waypoint generation, achieving strong results across diverse embodied tasks.
Xinyu Tang, Gangqiang Cao, Yurou Liu +13 authors
Scaling zero-reinforcement-learning to 1 trillion parameters improves reasoning quality and reveals emergent cognitive behaviors, supported by optimized training and structured evaluation.
Jiangnan Li, Yuqing Li, Mo Yu +2 authors
RARG integrates relevance scores into direct corpus search to prioritize documents and excerpts, accelerating accurate evidence retrieval for complex questions.
Haoyu Zhao, Xingyue Zhao, Siteng Huang +3 authors
A multi-modal 4D world model generates synchronized RGB, depth, and optical flow data from single RGB-D images and language instructions, enabling efficient robotic manipulation through unified diffusion processes and inverse dynamics policy learning.
Xinyu Geng, Xuanhua He, Sixiang Chen +7 authors
Training tool-use agents to improve from their own experience remains challenging, as supervised fine-tuning relies on fixed teacher-distilled trajectories, while sparse-reward reinforcement learning provides weak supervision for long-horizon interactions. We present DeepSearch-Evolve, a self-distillation framework for web agents built on DeepSearch-World, a deterministic and verifiable environment with reproducible search and page-reading tools. DeepSearch-World contains 420K multi-hop QA tasks constructed from entity-level random walks and supports key agentic cognitive behaviors useful for self-evolving, including progress verification, grounded reflection, and failure recovery. DeepSearch-Evolve iteratively performs trajectory generation, filtering, data mixing, and fine-tuning to train stronger agents. Without distillation from more capable models, DeepSearch-World-9B achieves competitive performance compared with open-source agents, reaching 31.2% on BrowseComp, 61.5% on GAIA, and 93.4% on HotpotQA, showing that verifiable environments enable scalable self-evolution for long-horizon web agents. We will release the environment, 420K training pool, validation set, model, and code to facilitate future research on self-improving deep search agents.
Jiaxing Li, Kai Zou, Cindy Zhou +7 authors
Joint distillation aligns initialization and refinement distributions to improve coverage and diversity in autoregressive video generation.
Qing Zong, Yue Guo, Mengxin Yang +2 authors
EvolvingWorld is a framework and benchmark that models interactive literary worlds as long-horizon co-evolution processes using coupled character and world modules with LLM-based state maintenance and evaluation.
AlayaWorld Team, Kaipeng Zhang, Chuanhao Li +14 authors
AlayaWorld is an open-source framework for creating interactive generative worlds that enables real-time user interaction and supports diverse actions within a modular architecture.
Letian Wang, Chuhan Zhang, Rishabh Kabra +9 authors
GenCeption uses a video generative diffusion backbone to build a generalist vision model that achieves strong performance across diverse perception tasks with high data efficiency and emergent generalization.
Chen Tang, Yizhou Wang, Jianyu Wu +26 authors
SciReasoner is a multimodal scientific foundation model that enables interpretable structural reasoning across proteins, molecules, and crystals by discretizing structural elements into a unified vocabulary for enhanced prediction and scientific inference.
Haozhe Wang, Weijia Feng, Jinpeng Yu +8 authors
Researchers introduce benchmarks and a co-training framework to enable search-augmented visual generation that addresses open-ended world-knowledge gaps.
Runhui Huang, Qihui Zhang, Zhe Liu +3 authors
SpectraReward leverages pretrained multimodal language models as training-free reward models for image-generation reinforcement learning by measuring prompt recovery likelihood, with a self-improving variant that uses the policy's own understanding branch to close the loop.
Jiayi Tian, Shiao Liu, Yuting Xu +30 authors
ABot-AgentOS is a general robotic agent operating system that provides deliberative planning, isolated skill execution, verification, and persistent multi-modal memory for long-horizon embodied tasks, evaluated on the EmbodiedWorldBench benchmark.
Xiang Hu, Xinyu Wei, Hao Gu +10 authors
Hierarchical Landmark Sparse Attention enables efficient long-context language modeling by learning chunk selection end-to-end, achieving performance comparable to full attention while extrapolating beyond training context lengths.
Bo-Wen Zhang, Junwei He, Wen Wang +5 authors
CoRT improves rubric-based reinforcement learning by using counterfactual replay to assign token-level credit weights within responses, boosting performance without auxiliary scorers.
Junlin Liu, Jiangwang Chen, Zixin Song +7 authors
Multi-Agent Protocol Distillation improves agentic search by using structured, style-normalized protocols to transfer reasoning from proprietary teachers to smaller models via dense distillation combined with sparse reinforcement learning.
Guixu Lin, Zheng-Hui Huang, Siqi Yang +3 authors
AlayaRenderer-Flash accelerates a generative world renderer to real-time speeds via few-step autoregressive streaming and distilled codecs while preserving structured scene dynamics.
Yuhang Wang, Yuling Shi, Shaoqiu Zhang +6 authors
SWE-Pruner Pro uses an agent's internal representations to prune tool outputs directly, reducing token usage while maintaining or improving coding task performance.
Yunxin Li, Jinchao Li, Shibo Su +7 authors
KnowAct-GUIClaw introduces a Know-Route-Act-Reflect framework that enhances cross-platform GUI automation and recursive self-improvement through experience-driven memory and skill libraries.
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