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。
Siming Huang, Tianhao Cheng, Jason Klein Liu +16 authors
OpenCoder is a top-tier open-access code LLM that provides comprehensive reproducible resources, including data, pipelines, and protocols, to advance code AI research.
50 篇论文 · 按点赞排序
Jiejun Tan, Zhicheng Dou, Wen Wang +3 authors
HtmlRAG enhances Retrieval-Augmented Generation (RAG) systems by using HTML instead of plain text, improving knowledge modeling and reducing information loss through HTML cleaning, compression, and pruning.
David Junhao Zhang, Roni Paiss, Shiran Zada +7 authors
ReCapture generates new videos with different camera trajectories from a user-provided video using multiview diffusion models and masked video fine-tuning.
Hongyu Wang, Shuming Ma, Furu Wei
BitNet a4.8 enhances the efficiency of large language models through 4-bit quantization and sparsification, achieving equivalent performance to BitNet b1.58 with reduced inference costs.
Antoine Grosnit, Alexandre Maraval, James Doran +15 authors
Agent K v1.0 automates data science tasks across various domains using structured reasoning and Bayesian optimization, achieving performance comparable to Expert-level Kaggle users.
Wenqiang Sun, Shuo Chen, Fangfu Liu +4 authors
DimensionX leverages video diffusion with spatial-temporal decoupling and trajectory-aware mechanisms to generate high-quality 3D and 4D scenes from single images.
Eldar Kurtic, Alexandre Marques, Shubhra Pandit +2 authors
A comprehensive study evaluates quantization formats for LLMs, showing that FP8 W8A8-FP is lossless, W8A8-INT incurs minimal accuracy loss, and W4A16-INT is cost-efficient, providing guidelines for optimal deployment.
Weixin Liang, Lili Yu, Liang Luo +8 authors
Mixture-of-Transformers (MoT) reduces pretraining computational costs for multi-modal large language models by decoupling non-embedding parameters by modality, enabling efficient processing and competitive performance across text, image, and speech.
Dingjie Song, Sicheng Lai, Shunian Chen +2 authors
A framework called MM-Detect is introduced to detect data contamination in multimodal large language models, demonstrating sensitivity across different contamination levels and identifying potential contamination stages during pre-training and fine-tuning.
Zhiyong Wu, Zhenyu Wu, Fangzhi Xu +8 authors
OS-Atlas improves GUI agent performance through a large open-source GUI grounding dataset and model innovations.
Yifan Xu, Xiao Liu, Xueqiao Sun +7 authors
AndroidLab provides a systematic framework for training and evaluating Android agents, supporting both large language models and multimodal models, and improves their task success rates.
Dang Nguyen, Viet Dac Lai, Seunghyun Yoon +9 authors
A new LLM agent framework enables dynamic action creation and reuse by generating and executing programs, offering greater flexibility and improved performance in complex, real-world scenarios.
Zehan Qi, Xiao Liu, Iat Long Iong +10 authors
WebRL enhances open LLMs as web agents through a self-evolving curriculum, robust outcome-supervised reward model, and adaptive reinforcement learning strategies, achieving high success rates compared to proprietary models.
Bingyi Kang, Yang Yue, Rui Lu +5 authors
Diffusion-based video generation models demonstrate perfect in-distribution generalization and measurable scaling behavior for combinatorial generalization but fail in out-of-distribution scenarios, prioritizing color over other factors.
Zhehao Zhang, Ryan A. Rossi, Branislav Kveton +18 authors
A taxonomy is introduced to bridge the gap between personalized text generation and personalization in LLMs for downstream applications, providing a comprehensive view of personalization techniques, datasets, and challenges.
Jaemin Cho, Debanjan Mahata, Ozan Irsoy +2 authors
M3DocRAG, a multi-modal RAG framework, enhances document visual question answering by handling multiple documents and modalities, achieving superior performance over existing methods.
Zhijian Zhuo, Ya Wang, Yutao Zeng +3 authors
Polynomial composition activations (PolyCom) enhance the representational capacity of transformer models, leading to improved approximation rates and performance in large language models.
Wenhao Wang, Yi Yang
TIP-I2V is a large-scale dataset of user-provided text and image prompts specifically for image-to-video generation, enabling advances in model development and safety.
Xingwu Sun, Yanfeng Chen, Yiqing Huang +103 authors
Hunyuan-Large, a large-scale Transformer-based mixture of experts model, outperforms smaller models across various benchmarks and provides insights into the scaling laws and learning rates for such models.
Anthony Chen, Jianjin Xu, Wenzhao Zheng +5 authors
Regional prompting is implemented for the FLUX diffusion model using attention manipulation to achieve fine-grained text-to-image generation without training.
David M. Chan, Rodolfo Corona, Joonyong Park +3 authors
Analysis of transformer-based models like LLaVA and Chameleon reveals similarities and differences in statistical behavior between discrete visual languages and natural languages, highlighting directions for improving computer vision models.
Muyang Li, Yujun Lin, Zhekai Zhang +7 authors
By quantizing the weights and activations of diffusion models to 4 bits using the SVDQuant paradigm and integrating a fused inference engine, memory usage and speed are significantly improved without compromising image quality.
Wei Cheng, Juncheng Mu, Xianfang Zeng +8 authors
MVPaint, a 3D texturing framework, addresses local discontinuities and multi-view inconsistencies in Text-to-Texture generation by employing synchronized multi-view generation, spatial-aware 3D inpainting, and UV refinement techniques, achieving high-fidelity textures.
Siddhesh Pawar, Junyeong Park, Jiho Jin +7 authors
Research investigates methods to incorporate cultural awareness into large language models, focusing on dataset creation, task strategies, benchmarking, and ethical considerations.
Dogyun Park, Sojin Lee, Sihyeon Kim +3 authors
A novel framework called Constant Acceleration Flow (CAF) improves few-step image generation by incorporating acceleration into the ODE flow model, demonstrating superior performance compared to existing methods.
Shehan Munasinghe, Hanan Gani, Wenqi Zhu +4 authors
VideoGLaMM, a Large Multimodal Model, achieves precise pixel-level grounding in videos using a dual vision encoder, spatio-temporal decoder, and tunable Visual-Language adapters, outperforming existing approaches on various multimodal tasks.
Kumara Kahatapitiya, Haozhe Liu, Sen He +4 authors
Adaptive Caching and Motion Regularization accelerate video diffusion transformers by optimizing computational resources based on video content and denoising needs.
Young-Jun Lee, Dokyong Lee, Junyoung Youn +2 authors
A new dataset and family of large language models, Thanos, improve conversational skills and quality of responses in social dialogue by infusing multifaceted conversational skills.
Jinyoung Park, Minseong Bae, Dohwan Ko +1 authors
LLaMo, a large language model for molecular graphs, uses a multi-level graph projector and cross-attention to achieve top performance in tasks like molecular description, property prediction, and IUPAC name generation.
Jonathan Roberts, Kai Han, Samuel Albanie
Research evaluates the information retrieval and reasoning capabilities of 17 long-context LLMs, revealing that many can handle multiple threads of information efficiently within a limited effective context window.
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