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。
Kevin Qinghong Lin, Linjie Li, Difei Gao +6 authors
ShowUI is a vision-language-action model that enhances GUI assistants by using UI-guided token selection and interleaved vision-language-action streaming, achieving high accuracy and efficiency in zero-shot screenshot grounding across different environments.
50 篇论文 · 按点赞排序
Yuchao Gu, Yipin Zhou, Yunfan Ye +5 authors
ROICtrl, an adapter for diffusion models using ROI-Align and ROI-Unpool, improves regional instance control in text-based visual generation with lower computational costs.
Nathan Lambert, Jacob Morrison, Valentina Pyatkin +20 authors
T\"ULU 3, an open-source family of post-trained language models, introduces transparent training data, recipes, and advanced techniques to match or surpass proprietary models in performance.
Zhenxiong Tan, Songhua Liu, Xingyi Yang +2 authors
OminiControl, a parameter-efficient framework, integrates image conditions into Diffusion Transformer models using a parameter reuse mechanism, excelling in subject-driven and spatially-aligned conditional generation.
Rundi Wu, Ruiqi Gao, Ben Poole +4 authors
CAT4D is a method that uses a multi-view video diffusion model to generate 4D dynamic scenes from monocular video by optimizing a deformable 3D Gaussian representation.
Shantanu Acharya, Fei Jia, Boris Ginsburg
Star Attention improves inference efficiency of large language models on long sequences by using block-sparse approximation, reducing memory and time without significant accuracy loss.
Xin Huang, Tengfei Wang, Ziwei Liu +1 authors
Material Anything is a unified diffusion framework that generates physically-based materials for 3D objects using a pre-trained image diffusion model with enhancements for stability and quality.
Zhen Huang, Haoyang Zou, Xuefeng Li +7 authors
Knowledge distillation from O1 model combined with fine-tuning achieves superior performance on complex mathematical reasoning and generalizes to open-ended QA tasks, promoting transparency in AI research.
Jooyoung Choi, Chaehun Shin, Yeongtak Oh +2 authors
The Style-friendly SNR sampler modifies the noise level distribution during fine-tuning to improve style alignment in diffusion models, enabling better capture of unique artistic styles.
Dawei Li, Bohan Jiang, Liangjie Huang +10 authors
A survey of LLM-based judgment and assessment explores definitions, taxonomy, benchmarks, challenges, and future directions for the LLM-as-a-judge paradigm.
Tianbin Li, Yanzhou Su, Wei Li +15 authors
GMAI-VL, a vision-language model trained on the GMAI-VL-5.5M dataset, achieves top performance in multimodal medical tasks by integrating visual and textual information.
Chaehun Shin, Jooyoung Choi, Heeseung Kim +1 authors
Diptych Prompting is a zero-shot text-to-image generation method that enhances subject alignment and attention between panels during inpainting, resulting in visually preferable images across various applications.
Shenghai Yuan, Jinfa Huang, Xianyi He +5 authors
ConsisID is a tuning-free, DiT-based IPT2V model that uses frequency-aware identity control to generate high-quality, identity-preserving videos by decomposing facial features into low and high-frequency elements.
Noam Rotstein, Gal Yona, Daniel Silver +3 authors
Image editing is improved using pretrained video models to ensure accurate edits and preserve image fidelity by reformulating the process as a smooth temporal transition.
Di Zhang, Jingdi Lei, Junxian Li +10 authors
The Critic-V framework leverages an Actor-Critic paradigm to enhance multimodal reasoning in vision-language models by decoupling reasoning and critique processes, significantly improving performance and reliability.
Chaoyun Zhang, Shilin He, Jiaxu Qian +9 authors
LLM-brained GUI agents enable sophisticated interactions through natural language commands, automating tasks across various platforms and enhancing user experiences.
Yoel Zimmermann, Adib Bazgir, Zartashia Afzal +138 authors
The hackathon showcased the diverse applications of large language models in materials science and chemistry, highlighting improvements in prediction, design, automation, communication, data management, hypothesis evaluation, and knowledge extraction.
Duong H. Le, Tuan Pham, Sangho Lee +5 authors
OneDiffusion is a versatile diffusion model that supports bidirectional image synthesis and understanding, enabling tasks like conditional generation, image deblurring, and camera pose estimation through unified frame sequence training and conditioning.
Riza Velioglu, Petra Bevandic, Robin Chan +1 authors
TryOffDiff, an adaptation of Stable Diffusion for generating standardized garment images from dressed individuals, outperforms pose transfer and virtual try-on methods using SigLIP-based visual conditioning and the DISTS metric for evaluation.
Sankalp Sinha, Mohammad Sadil Khan, Muhammad Usama +4 authors
A novel dataset MARVEL-40M+ with extensive text annotations and a two-stage pipeline for high-fidelity 3D content generation from text prompts outperforms existing datasets in annotation quality and diversity.
Shaohan Huang, Xun Wu, Shuming Ma +1 authors
A novel implementation of Multi-Head Mixture-of-Experts maintains efficiency and surpasses traditional models in language performance, even with 1-bit Large Language Models.
Chengyou Jia, Changliang Xia, Zhuohang Dang +3 authors
ChatGen-Evo, a multi-stage evolution strategy, automates text-to-image generation by improving accuracy and image quality through systematic evaluation on ChatGenBench.
Luchao Qi, Jiaye Wu, Bang Gong +3 authors
MyTimeMachine combines global aging models with personal photo collections to generate realistic, identity-preserving aging images and videos.
Junlong Cheng, Bin Fu, Jin Ye +10 authors
The IMed-361M benchmark dataset, featuring over 6.4 million medical images and 361 million masks, enhances interactive medical image segmentation through a foundation model and diverse, high-density annotations.
Gabriel Chua, Shing Yee Chan, Shaun Khoo
A flexible methodology for creating off-topic guardrails in LLMs uses synthetic datasets generated by LLMs to improve accuracy and generalization over heuristic approaches.
Chaoyou Fu, Yi-Fan Zhang, Shukang Yin +9 authors
A comprehensive survey of evaluation methods and benchmarks for Multimodal Large Language Models (MLLMs) to improve model performance and drive research progress.
Dongping Chen, Ruoxi Chen, Shu Pu +8 authors
ISG provides a comprehensive evaluation framework and benchmark for interleaved text-and-image generation, highlighting challenges and improvements over unified and compositional models.
Yuhang Han, Xuyang Liu, Pengxiang Ding +4 authors
A unified filter-correlate-compress paradigm accelerates inference in Multimodal Large Language Models with minimal impact on performance.
Jin Ye, Ying Chen, Yanjun Li +7 authors
A large-scale benchmark evaluates the transferability of full-body CT pre-trained models across different medical image modalities and segmentation tasks, demonstrating effective modality and target adaptability.
Kaichen Zhang, Yifei Shen, Bo Li +1 authors
A framework using Sparse Autoencoders and automatic interpretation helps understand human-understandable features in Large Multimodal Models (LMMs), providing insights into model behavior and cognitive processes.
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