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
Wenhao Yan, Fengjia Guo, Zhuoyi Yang +1 authors
SCAIL-2 enables end-to-end character animation by directly transferring motion from driving videos without intermediate representations, using unified task decomposition and synthetic data generation.
Wei Pang, Xiangru Jian, Hehan Li +10 authors
TRL-Bench establishes a standardized benchmark for evaluating tabular representation learning models across multiple granularities, revealing that encoder performance varies by task type and requires capability-specific assessment rather than single leaderboard rankings.
Jiangtao Wu, Jiaming Wang, Yiwen He +7 authors
A new benchmark called CoVEBench is introduced to evaluate compositional video editing capabilities, addressing limitations of existing models in handling complex, multi-step editing tasks while preserving spatiotemporal content.
Daria Ledneva, Mikhail Nuridinov, Denis Kuznetsov
GENEB presents a comprehensive benchmark for evaluating genomic foundation models across diverse tasks and architectures under a unified protocol.
Hongcheng Gao, Hailong Qu, Jingyi Tang +18 authors
SpatialWorld presents a unified benchmark for evaluating interactive spatial understanding in multimodal agents through diverse real-world tasks with partial observability and text-based actions.
Ziyang Ma, Ruiqi Yan, Ruiyang Xu +35 authors
MMAE presents a comprehensive benchmark for instruction-based audio editing across multiple modalities and complexity levels, revealing significant gaps in current model capabilities.
Bowen Ping, Xiangxin Zhou, Penghui Qi +3 authors
Flow-DPPO replaces ratio clipping with divergence proximal constraints in flow matching models, improving training stability and multi-objective optimization through exact KL divergence computation.
Cong Chen, Guo Gan, Kaixiang Ji +7 authors
MemDreamer addresses long-video understanding challenges by decoupling perception and reasoning through hierarchical graph memory and agentic exploration, achieving state-of-the-art performance with reduced computational overhead.
Yuchen Xian, Yang He, Yunqiu Xu +1 authors
VIA-SD introduces a multi-tier speculative decoding framework that uses intra-model routing to reduce verification costs by employing slim submodels for medium-confidence token validation, achieving significant speedups over traditional approaches.
Jiale Zhao, Guoxin Chen, Fanzhe Meng +4 authors
A large-scale dataset called DeNovoSWE is introduced for training code agents to generate entire software repositories from documentation, significantly improving performance on long-horizon software engineering tasks.
Paul Hyunbin Cho, Jinhyuk Jang, SeokYoung Lee +7 authors
Autoregressive diffusion method for video-to-video lip synchronization achieves real-time performance through distillation and optimized inference schedules.
Woojung Song, Dongmin Choi, Yoonah Park +3 authors
Human psychometric questionnaires fail to reliably predict LLM behavior in real-world interactions, while generation-based profiling offers superior accuracy for understanding model responses to everyday user queries.
Jiarui Yao, Xiangxin Zhou, Penghui Qi +3 authors
DRPO improves LLM reinforcement learning stability by replacing hard masks with smooth regularization that provides continuous gradient corrections beyond trust-region boundaries.
Amy Xin, Jiening Siow, Junjie Wang +5 authors
Environment engineering enhances autonomous scientific discovery by designing structured agent environments that optimize behaviors like exploration and collaboration while mitigating issues such as reward hacking and human oversight friction, as demonstrated by the EurekAgent system that achieves state-of-the-art results across multiple domains with low computational costs.
Yuchen Xian, Yunqiu Xu, Yang He +1 authors
A multimodal image fusion approach uses a 1D token interface from a pretrained image tokenizer to enhance global appearance coherence while preserving local details through selective token editing.
Yuke Zhao, Wangbo Zhao, Weijie Wang +8 authors
WorldOlympiad presents a comprehensive benchmark for evaluating video-based world models across physical faithfulness, geometric consistency, and interaction fidelity, revealing significant gaps in current generative models' capabilities.
Wayne King, Zeyue Xue, Yuxuan Bian +13 authors
Controlled study of memory mechanisms in action-conditioned world models reveals that memory structure and capacity significantly impact open-domain return performance beyond simple replay fidelity measures.
Guozhen Zhang, Xuerui Qiu, Yutao Cui +11 authors
HYDRA-X presents a unified multimodal model that integrates image and video tokenization within a single Vision Transformer, addressing spatiotemporal reconstruction and semantic awareness through causal temporal attention and hierarchical compression.
Chaofan Ma, Zhenjie Mao, Yuhuan Yang +5 authors
A training-free framework for spatial reasoning from egocentric videos that enables revisiting conclusions through synthesized novel-view videos generated from predicted 3D geometry.
Elis Karcini, Faisal Mehrban, Quang Nguyen +6 authors
Robot intelligence advancement requires integrating unstructured behavioral data through specialized interfaces for labeling, embodiment mapping, world modeling, and reward inference rather than relying solely on policy scaling.
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