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。
Ahmed Nassar, Andres Marafioti, Matteo Omenetti +10 authors
SmolDocling is a compact vision-language model that performs end-to-end document conversion with robust performance across various document types using 256M parameters and a new markup format.
50 篇论文 · 按点赞排序
Bo Peng, Ruichong Zhang, Daniel Goldstein +12 authors
RWKV-7 "Goose" achieves state-of-the-art performance in multilingual tasks with optimal memory and inference efficiency, exceeding Transformer capabilities in complexity.
Qiying Yu, Zheng Zhang, Ruofei Zhu +32 authors
The DAPO algorithm, which includes decoupled clip and dynamic sampling policy optimization, enables open-source, high-performance reinforcement learning training for large-scale language models, enhancing reproducibility in the field.
Jianhong Bai, Menghan Xia, Xiao Fu +8 authors
ReCamMaster uses pre-trained text-to-video models to render dynamic scenes of input videos from novel camera trajectories, leveraging a unique video conditioning mechanism and a custom dataset created with Unreal Engine 5.
Runze Zhang, Guoguang Du, Xiaochuan Li +10 authors
The DropletVideo model preserves spatio-temporal coherence in generated videos by considering both camera movements and plot developments, leveraging a large annotated dataset.
Asaf Yehudai, Lilach Eden, Alan Li +5 authors
This survey analyzes evaluation methodologies for large language model-based agents, covering fundamental capabilities, application-specific benchmarks, and generalist agents, highlighting trends and gaps in the field.
Daniil Selikhanovych, David Li, Aleksei Leonov +6 authors
RSD is a distillation method for ResShift, a top diffusion-based super-resolution model, that achieves high-quality results with fewer parameters and GPU memory compared to other methods.
Kwanyoung Kim, Byeongsu Sim
PLADIS leverages sparse attention in cross-attention layers to enhance pre-trained text-to-image diffusion models, improving text alignment and human preference without additional training.
Yang Sui, Yu-Neng Chuang, Guanchu Wang +9 authors
This survey examines progress and methods for improving efficient reasoning in Large Language Models (LLMs), focusing on model, output, and input prompt optimizations, as well as the use of efficient data and small models.
Haoqi Yuan, Yu Bai, Yuhui Fu +6 authors
Being-0, a hierarchical framework integrating Foundation Models with a modular skill library and a lightweight vision-language model, achieves efficient real-time performance on complex humanoid robot tasks by coordinating high-level cognition with low-level control.
Zechen Bai, Hai Ci, Mike Zheng Shou
IPV-Bench evaluates video generation and understanding models on creating and interpreting impossible videos, highlighting their limitations and guiding future advancements.
Zorik Gekhman, Eyal Ben David, Hadas Orgad +5 authors
LLMs encode more internal factual knowledge than they express externally, with some knowledge so deeply hidden that it is never generated, despite repeated sampling.
Quy-Anh Dang, Chris Ngo
Reinforcement learning enhances reasoning in small language models with limited resources, demonstrating significant performance improvements in mathematical reasoning tasks.
NVIDIA, Alisson Azzolini, Hannah Brandon +42 authors
Cosmos-Reason1 models, using hierarchical and two-dimensional ontologies for physical common sense and embodied reasoning, generate embodied decisions through multimodal large language models trained in vision and Physical AI stages.
Fangzhi Xu, Hang Yan, Chang Ma +4 authors
A new decoding strategy, $\phi$-Decoding, uses foresight sampling and pruning techniques to enhance the efficiency and performance of large language models during inference.
Mert Cemri, Melissa Z. Pan, Shuyi Yang +10 authors
A comprehensive study identifies 14 unique failure modes in Multi-Agent Systems, categorizing them into specification issues, inter-agent conflicts, and task verification problems, and proposes interventions for preventing these failures.
Zineng Tang, Long Lian, Seun Eisape +6 authors
TULIP enhances image-text contrastive models by incorporating generative data augmentation and contrastive learning to improve fine-grained visual features and zero-shot performance.
Xinyu Fang, Zhijian Chen, Kai Lan +9 authors
Creation-MMBench evaluates the creative capabilities of Multimodal Large Language Models (MLLMs) in image-based real-world tasks, revealing performance gaps and the negative impacts of visual fine-tuning.
Ruowen Zhao, Junliang Ye, Zhengyi Wang +4 authors
DeepMesh generates high-quality 3D meshes with intricate details and precise topology using an efficient pre-training strategy and reinforcement learning, outperforming existing methods.
Haoran Feng, Zehuan Huang, Lin Li +2 authors
Personalize Anything is a training-free framework for personalized image generation using diffusion transformers, enabling subject consistency and flexibility through timestep-adaptive token replacement and patch perturbation.
Zeqiang Lai, Yunfei Zhao, Zibo Zhao +10 authors
FlashVDM accelerates both diffusion sampling and VAE decoding in 3D diffusion models, achieving high performance with reduced inference times.
Dewei Zhou, Mingwei Li, Zongxin Yang +1 authors
DreamRenderer enhances image-conditioned generation by precisely controlling instance attributes using bounding boxes and masks, improving the Image Success Ratio and layout-to-image performance.
Nikita Starodubcev, Denis Kuznedelev, Artem Babenko +1 authors
A scale-wise distillation framework for diffusion models reduces computational costs and improves inference times by incorporating next-scale predictions and enhancing distribution matching methods.
Muyao Li, Zihao Wang, Kaichen He +2 authors
A novel self-supervised post-training method for Visual Language Models (VLMs) enhances their performance in action-based decision-making tasks, achieving state-of-the-art results in Minecraft.
Jianyuan Wang, Minghao Chen, Nikita Karaev +3 authors
VGGT, a feed-forward neural network, efficiently infers multiple 3D attributes from single or multiple views, outperforming alternatives and enhancing downstream tasks without post-processing.
Liming Jiang, Qing Yan, Yumin Jia +3 authors
InfiniteYou leverages Diffusion Transformers to generate high-fidelity identity-preserved images, addressing issues of identity similarity, text-image alignment, and quality through a novel framework and training strategy.
Chaoyun Zhang, Shilin He, Liqun Li +4 authors
The paper compares API-based and GUI-based LLM agents, analyzing their differences and potential integration to guide practitioners in selecting or combining these automation paradigms.
Yaoting Wang, Shengqiong Wu, Yuecheng Zhang +4 authors
A systematic survey of multimodal chain-of-thought reasoning, addressing foundational concepts, methodologies, challenges, and future directions for multimodal applications.
Jianzhu Yao, Kevin Wang, Ryan Hsieh +5 authors
SPIN-Bench evaluates strategic planning and social reasoning in AI through a unified framework combining PDDL tasks, board games, card games, and negotiation scenarios, highlighting challenges in deep multi-hop reasoning and social coordination.
Siyuan Huang, Yue Liao, Siyuan Feng +5 authors
Adversarial Data Collection, a Human-in-the-Loop framework that uses dynamic perturbations, reduces data needs while enhancing the performance and robustness of robotic manipulation 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