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。
Nikhila Ravi, Valentin Gabeur, Yuan-Ting Hu +15 authors
Segment Anything Model 2 (SAM 2) uses a transformer architecture with streaming memory to achieve high performance in image and video segmentation, requiring fewer interactions and faster processing than previous models.
50 篇论文 · 按点赞排序
Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey +530 authors
Llama 3, a multilingual and multi-modal language model with 405B parameters, achieves competitive performance across tasks including image, video, and speech recognition when integrated through a compositional approach.
Gemma Team, Morgane Riviere, Shreya Pathak +193 authors
Gemma 2 introduces improvements in the Transformer architecture through interleaving local-global attentions and group-query attention, showcasing superior performance relative to its size.
Leon Voukoutis, Dimitris Roussis, Georgios Paraskevopoulos +6 authors
Developers created Meltemi 7B, a 7 billion parameter open-source large language model for Greek, trained on a 40 billion token corpus, and enhanced with instruction-tuning for chat applications.
Pierre Colombo, Telmo Pires, Malik Boudiaf +7 authors
Two large legal language models, SaulLM-54B and SaulLM-141B, based on the Mixtral architecture, are introduced for domain-specific adaptation in the legal sector using continued pretraining, specialized protocols, and preference alignment with synthetic data.
Santosh V. Patapati
A BiLSTM-based multi-modal model combines audio, facial, and text data using a GPT-4 model to classify depression, achieving state-of-the-art performance in binary classification tasks.
Wenxuan Zhang, Hou Pong Chan, Yiran Zhao +9 authors
SeaLLMs 3 is a large language model optimized for Southeast Asian languages, offering state-of-the-art performance in various tasks while addressing safety and cultural considerations.
Yu Lu, Yuanzhi Liang, Linchao Zhu +1 authors
FreeLong enhances long video generation by balancing global and local video features during denoising, improving quality and consistency.
Jinghuan Shang, Karl Schmeckpeper, Brandon B. May +4 authors
Theia, a vision foundation model for robot learning, outperforms existing models with less data and smaller sizes by distilling knowledge from multiple vision foundation models.
Zehui Chen, Kuikun Liu, Qiuchen Wang +4 authors
MindSearch, an LLM-based multi-agent framework, improves web information seeking and integration through parallel processing and hierarchical retrieval, achieving better performance than existing solutions.
Aleksandar Shtedritski, Christian Rupprecht, Andrea Vedaldi
SHIC leverages foundation computer vision models to learn canonical surface maps without manual supervision, achieving superior results by simulating the annotation process using image-to-image correspondences and enhanced template views.
Guoli Yin, Haoping Bai, Shuang Ma +21 authors
MMAU benchmark evaluates large language models' capabilities across multiple domains and essential skills using comprehensive offline tasks to enhance interpretability and reliability.
Gagan Jain, Nidhi Hegde, Aditya Kusupati +5 authors
Mixture of Nested Experts (MoNE) reduces inference time by dynamically processing tokens through nested experts, achieving performance equivalent to baseline models.
Wenxuan Wang, Quan Sun, Fan Zhang +3 authors
DIVA enhances CLIP's performance through a self-supervised diffusion process, improving visual capabilities and multimodal understanding without additional text labels.
Harsh Trivedi, Tushar Khot, Mareike Hartmann +6 authors
A comprehensive benchmark and execution environment for evaluating autonomous agents performing complex, realistic digital tasks across multiple apps through rich code generation.
Tianduo Wang, Shichen Li, Wei Lu
Enhancing small-scale language models for mathematical reasoning through self-training and preference learning, resulting in improved performance and reduced costs compared to large proprietary models.
Boyi Li, Ligeng Zhu, Ran Tian +17 authors
Wolf is a framework that combines Vision Language Models for efficient video captioning, outperforming existing approaches and introducing a quality assessment metric called CapScore.
Eduardo Soares, Victor Shirasuna, Emilio Vital Brazil +3 authors
A large-scale pre-trained encoder-decoder chemical language model achieves state-of-the-art performance across various tasks by leveraging a vast dataset of molecular structures, exhibiting strong few-shot learning capabilities.
Yuhui Xu, Zhanming Jie, Hanze Dong +6 authors
ThinK, a novel query-dependent KV cache pruning method, effectively reduces memory costs by over 20% in long-context scenarios without compromising the accuracy of Large Language Models (LLMs).
Mark Boss, Zixuan Huang, Aaryaman Vasishta +1 authors
SF3D quickly reconstructs high-quality textured 3D meshes from a single image by training for mesh generation with fast UV unwrapping, material parameter prediction, normal map estimation, and low-frequency illumination removal.
Aisyah Razak, Ariff Nazhan, Kamarul Adha +3 authors
A novel safe-for-work text classifier for Malaysian language content is developed using a curated dataset of Malaysian text and state-of-the-art natural language processing techniques, enhancing alignment in LLM-Ops.
Zhenghao Zhang, Junchao Liao, Menghao Li +2 authors
Tora, a trajectory-oriented Diffusion Transformer framework, integrates textual, visual, and trajectory conditions for high-fidelity video generation with controlled motion.
Zhenyu Tang, Junwu Zhang, Xinhua Cheng +5 authors
A Cycle3D framework uses a unified 2D diffusion and 3D reconstruction model to generate high-quality, consistent 3D content.
Xin Zhang, Yanzhao Zhang, Dingkun Long +10 authors
A long-context multilingual text representation model and reranker are developed and evaluated, showing superior performance and efficiency compared to existing models.
Yadong Lu, Jianwei Yang, Yelong Shen +1 authors
OmniParser enhances GPT-4V's interaction with user interfaces by accurately parsing screenshots into actionable elements, significantly improving performance on key benchmarks.
Trapoom Ukarapol, Zhicheng Lee, Amy Xin
Contrastive fine-tuning improves text embeddings and performance in smaller language models across various benchmarks.
Gili Goldin, Shuly Wintner
Knesset-DictaBERT, a Hebrew language model fine-tuned on Israeli parliamentary data, shows enhanced performance in understanding parliamentary language compared to the original DictaBERT.
Norman Di Palo, Leonard Hasenclever, Jan Humplik +1 authors
Diffusion Augmented Agents (DAAG) enhances reinforcement learning for embodied agents using large and vision language models, diffusion models, and hindsight experience augmentation to improve sample efficiency and transfer learning.
Benlin Liu, Yuhao Dong, Yiqin Wang +4 authors
A visual prompting method called Coarse Correspondence enhances 3D and temporal understanding in multimodal language models, achieving state-of-the-art results on relevant benchmarks.
Zhixiang Wang, Baiang Li, Jian Wang +4 authors
Latent diffusion models with pre-trained knowledge enhance image matting by generating high-quality, detailed, and photorealistic mattes.
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