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。
44 篇论文 · 按点赞排序
Lincong Feng, Muyu Wang, Maoyu Wang +2 authors
MetaDreammer, a two-stage optimization approach, integrates 2D and 3D prior knowledge to generate high-quality, multi-view consistent 3D objects from textual prompts efficiently.
Wen Wang, Canyu Zhao, Hao Chen +3 authors
The system uses large language models for layout planning and large-scale text-to-image models for generating high-quality, consistent story images with minimal human interaction, transforming bounding boxes into dense control conditions to improve image quality.
Mengyang Feng, Jinlin Liu, Miaomiao Cui +1 authors
A circular blending strategy is introduced to generate seamless 360-degree panoramic images using diffusion models for both text-to-360-panoramas and single-image-to-360-panoramas tasks.
Jiaxi Lv, Yi Huang, Mingfu Yan +6 authors
GPT4Motion integrates GPT, Blender, and Stable Diffusion to generate high-quality, motion-coherent videos from text prompts more efficiently than previous methods.
Silen Naihin, David Atkinson, Marc Green +5 authors
A safety framework with a context-sensitive monitor is proposed to conduct safe testing of autonomous agents in the real world, identifying and stopping unsafe behaviors through adversarial simulations and real-world tests.
Nicholas Farn, Richard Shin
ToolTalk is a benchmark for evaluating LLM-based assistants using complex user intents requiring multi-step tool interactions, with success rates of 26% and 50% for GPT-3.5 and GPT-4 respectively.
Yunshan Zhong, Jiawei Hu, Mingbao Lin +2 authors
I&S-ViT introduces methods to improve post-training quantization of Vision Transformers, specifically addressing quantization inefficiency and rugged loss landscapes to maintain performance in low-bit scenarios.
Yining Ye, Xin Cong, Shizuo Tian +9 authors
Agentic Process Automation (APA) employs LLM-based agents to handle intelligent workflow design and dynamic decision-making in automation, marking a new paradigm shift.
Hritik Bansal, Yonatan Bitton, Idan Szpektor +2 authors
A video-language alignment model finetuned with VideoCon, a dataset of plausible contrastive video captions, achieves superior performance in video-language tasks compared to existing models.
Yilun Kong, Jingqing Ruan, Yihong Chen +9 authors
A framework is introduced to improve the task planning and tool usage capabilities of LLM-based agents in real-world systems through API retrieval, LLM fine-tuning, and demonstration selection.
Lihan Zha, Yuchen Cui, Li-Heng Lin +5 authors
A large language model-based system, DROC, effectively processes and applies human corrections to improve robot performance in new environments by distilling and retrieving relevant knowledge.
Atin Sakkeer Hussain, Shansong Liu, Chenshuo Sun +1 authors
A framework integrating large language models for multi-modal music understanding and generation outperforms current state-of-the-art models.
Naoki Wake, Atsushi Kanehira, Kazuhiro Sasabuchi +2 authors
A pipeline integrating human action observations into a general-purpose vision-language model enables robotic manipulation through task planning and spatiotemporal grounding of affordance data.
Chris Kelly, Luhui Hu, Cindy Yang +6 authors
UnifiedVisionGPT is a novel framework that consolidates and automates the integration of state-of-the-art vision models, enhancing computer vision through efficiency, versatility, generalization, and performance.
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