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。
686 篇论文 · 按点赞排序
Shuqi Lu, Chaofan Li, Kun Luo +21 authors
AREX is a recursively self-improving deep research agent that verifies answers constraint-wise, compresses verified evidence into a compact state, and uses targeted follow-up research to refine results over long horizons.
Zhanhui Zhou, Lingjie Chen, Hanghang Tong +1 authors
A unified open-source framework is presented that standardizes core components of diffusion language modeling for reproduction, customization, and accessible development of both large and small models.
Jianyu Wu, Yizhou Wang, Encheng Su +2 authors
Dual-Anchored Policy Distillation resolves privilege illusion in on-policy self-distillation by aligning matched-information paths and reducing reliance on privileged guidance, improving performance across model scales.
Yukang Cao, Haozhe Xie, Fangzhou Hong +4 authors
HSImul3R presents a unified framework for 3D reconstruction of human-scene interactions that bridges the perception-simulation gap through physics-grounded bidirectional optimization and reinforcement learning.
Xunhao Lai, Weiqi Xu, Yufeng Yang +8 authors
MiniMax Sparse Attention enables efficient processing of ultra-long contexts in large language models through blockwise sparsity and optimized GPU execution, achieving significant speedups while maintaining performance.
Tingyun Li, Wenfeng Feng, Weiqing Li +3 authors
Boundary-Calibrated Intervention Transfer selectively reuses past training evidence by checking contextual applicability and running bounded trials, reducing harmful updates and improving final model quality in autonomous post-training.
Yuxin Zuo, Zikai Xiao, Li Sheng +30 authors
Language-based world models enable agentic environment simulation across multiple domains and enhance general agent performance through scalable simulation and improved downstream task performance.
Xinlei Yu, Zhangquan Chen, Yongbo He +34 authors
Latent space is emerging as a fundamental computational substrate for language-based models, offering advantages over explicit token-level approaches through continuous representation that mitigates linguistic redundancy and sequential inefficiency.
Jianuo Huang, Yaojie Zhang, Qituan Zhang +3 authors
Domino is a speculative decoding framework that improves LLM inference speed by decoupling causal dependency modeling from autoregressive drafting through a parallel backbone and lightweight causal refinement head, achieving significant speedups in both end-to-end execution and throughput.
Yulin Li, Tengyao Tu, Li Ding +5 authors
ReBalance is a training-free framework that balances reasoning in large models by using confidence indicators to detect and correct overthinking and underthinking behaviors through dynamic steering vectors.
Yixuan Wang, Yifei Chen, Haichao Zhang +6 authors
SA-MRPO independently standardizes multi-objective rewards and adaptively discounts saturated objectives to redirect optimization toward under-optimized goals.
Yiying Yang, Wei Cheng, Sijin Chen +5 authors
OmniLottie framework generates high-quality vector animations from multi-modal instructions using a specialized Lottie tokenizer and pretrained vision-language models.
Yi Ding, Ruqi Zhang
On-policy distillation relies mainly on suppressing low-probability tokens rather than teacher guidance, motivating a supervision-free entropy-adaptive method that substantially improves reasoning performance.
Zhiyuan Hu, Yucheng Wang, Yufei He +7 authors
Reinforcement learning for large language models is enhanced by a rollout-level objective that rewards rare high-level reasoning strategies, improving diverse solution discovery without sacrificing initial performance.
Han Luo, Bingbing Wen, Lucy Lu Wang
Agentic abstention involves determining when an AI agent should cease interaction under uncertainty, requiring sequential decision-making across multiple environments and task types.
Hongrui Jia, Chaoya Jiang, Shikun Zhang +1 authors
Diagnostic-driven Progressive Evolution enables continuous improvement of large multimodal models through iterative diagnosis and targeted data generation guided by identified weaknesses.
Meituan LongCat Team, Bin Xiao, Chao Wang +86 authors
Discrete Native Autoregressive framework enables unified multimodal processing by representing diverse modalities in a shared discrete space through a novel visual transformer architecture.
Mikhail Komarov, Ivan Bondarenko, Stanislav Shtuka +5 authors
RAGU is a modular GraphRAG engine that separates knowledge extraction from consolidation and uses a small language-optimized model to build cleaner graphs, improving retrieval and synthesis over existing systems.
Shiyuan Feng, Huan-ang Gao, Haohan Chi +7 authors
Direct On-Policy Distillation transfers reinforcement learning improvements from smaller to larger models by using the policy shift induced by RL as an implicit reward signal, enabling efficient scaling of training without re-running expensive RL on the target model.
Xin Ding, Liang Mi, Mingzhe Huang +12 authors
Zetta is a closed-loop embodied harness that evolves runtime critics and recovery skills online to govern physical execution at action frequency, achieving high success on robot benchmarks with faster inference and scaling self-exploration.
Yuwen Du, Rui Ye, Shuo Tang +4 authors
Deep search capabilities have become an indispensable competency for frontier Large Language Model (LLM) agents, yet the development of high-performance search agents remains dominated by industrial giants due to a lack of transparent, high-quality training data. This persistent data scarcity has fundamentally hindered the progress of the broader research community in developing and innovating within this domain. To bridge this gap, we introduce OpenSeeker, the first fully open-source search agent (i.e., model and data) that achieves frontier-level performance through two core technical innovations: (1) Fact-grounded scalable controllable QA synthesis, which reverse-engineers the web graph via topological expansion and entity obfuscation to generate complex, multi-hop reasoning tasks with controllable coverage and complexity. (2) Denoised trajectory synthesis, which employs a retrospective summarization mechanism to denoise the trajectory, therefore promoting the teacher LLMs to generate high-quality actions. Experimental results demonstrate that OpenSeeker, trained (a single training run) on only 11.7k synthesized samples, achieves state-of-the-art performance across multiple benchmarks including BrowseComp, BrowseComp-ZH, xbench-DeepSearch, and WideSearch. Notably, trained with simple SFT, OpenSeeker significantly outperforms the second-best fully open-source agent DeepDive (e.g., 29.5% v.s. 15.3% on BrowseComp), and even surpasses industrial competitors such as Tongyi DeepResearch (trained via extensive continual pre-training, SFT, and RL) on BrowseComp-ZH (48.4% v.s. 46.7%). We fully open-source the complete training dataset and the model weights to democratize frontier search agent research and foster a more transparent, collaborative ecosystem.
Yining Chen, Jihao Zhao, Bo Tang +3 authors
MemPrivacy enables privacy-preserving personalized memory in edge-cloud environments by using type-aware placeholders to protect sensitive data while maintaining semantic integrity for effective memory operations.
Shiva Krishna Reddy Malay, Shravan Nayak, Jishnu Sethumadhavan Nair +6 authors
Large language models are shifting from passive information providers to active agents intended for complex workflows. However, their deployment as reliable AI workers in enterprise is stalled by benchmarks that fail to capture the intricacies of professional environments, specifically, the need for long-horizon planning amidst persistent state changes and strict access protocols. In this work, we introduce EnterpriseOps-Gym, a benchmark designed to evaluate agentic planning in realistic enterprise settings. Specifically, EnterpriseOps-Gym features a containerized sandbox with 164 database tables and 512 functional tools to mimic real-world search friction. Within this environment, agents are evaluated on 1,150 expert-curated tasks across eight mission-critical verticals (including Customer Service, HR, and IT). Our evaluation of 14 frontier models reveals critical limitations in state-of-the-art models: the top-performing Claude Opus 4.5 achieves only 37.4% success. Further analysis shows that providing oracle human plans improves performance by 14-35 percentage points, pinpointing strategic reasoning as the primary bottleneck. Additionally, agents frequently fail to refuse infeasible tasks (best model achieves 53.9%), leading to unintended and potentially harmful side effects. Our findings underscore that current agents are not yet ready for autonomous enterprise deployment. More broadly, EnterpriseOps-Gym provides a concrete testbed to advance the robustness of agentic planning in professional workflows.
Liangcai Su, Zhaopeng Feng, Zhuo Chen +13 authors
FrontierChallenge evaluates end-to-end scientific workflows across domains, revealing that frontier models complete only about 20% of tasks despite high partial scores and frequent claims of completion.
Weimin Xiong, Shuhao Gu, Bowen Ye +5 authors
A large-scale GUI dataset was created by automatically extracting interaction trajectories from internet videos, enabling improved performance in GUI agents through pre-training on this diverse collection.
Junchao Huang, Guian Fang, Shengju Qian +15 authors
SolarWM provides an open framework and unified training recipe for building interactive video world models across diverse data sources and generator backbones, enabling long-horizon real-time rollouts.
Mohamad Zbib, Mohamad Bazzi, Ammar Mohanna +2 authors
Speculative decoding effectiveness depends on draft model training data alignment with downstream tasks, with specialized drafters performing better when combined through confidence-based routing rather than simple averaging.
Yuandong Pu, Le Zhuo, Sayak Paul +11 authors
The study formalizes probabilistic alignment for world models, introduces PAWBench and PAWEval to evaluate video generators as stochastic samplers, and finds current models fail to match reference behavior distributions.
Qiuyue Wang, Mingsheng Li, Jian Guan +37 authors
A unified vision-language-action model is presented that integrates diverse embodied decision-making tasks through a shared architecture and training approach, demonstrating strong performance across manipulation, navigation, and trajectory prediction with generalization across different robot platforms and environments.
Shijie Lian, Bin Yu, Xiaopeng Lin +10 authors
PhysBrain 1.0 leverages human egocentric video to generate physical commonsense supervision for vision-language-action models, achieving state-of-the-art performance in embodied control tasks through capability-preserving adaptation.
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