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。
Seungone Kim, Jamin Shin, Yejin Cho +8 authors
Prometheus, an open-source LLM, matches GPT-4's evaluation capabilities using customized score rubrics and outperforms ChatGPT across various benchmarks.
9 篇论文 · 按点赞排序
Jingkang Yang, Yuhao Dong, Shuai Liu +8 authors
Octopus, a novel vision-language model, integrates multimodal perception and reasoning to perform diverse tasks using GPT-4 and Reinforcement Learning with Environmental Feedback in an experimental environment called OctoVerse.
Yiheng Xu, Hongjin Su, Chen Xing +13 authors
Lemur and Lemur-Chat are open-source language models combining language and coding capabilities, achieving state-of-the-art performance across various benchmarks and agent tasks.
Taoran Yi, Jiemin Fang, Guanjun Wu +5 authors
A framework combining 2D and 3D diffusion models using 3D Gaussian splatting efficiently generates high-quality 3D assets from text prompts with real-time rendering.
Zhengyuan Yang, Jianfeng Wang, Linjie Li +4 authors
Idea to Image uses iterative multimodal refinement with GPT-4V to enhance text-to-image generation by refining prompts based on model characteristics, leading to higher-quality images.
Xian Liu, Jian Ren, Aliaksandr Siarohin +6 authors
A unified framework, HyperHuman, generates hyper-realistic human images using a Latent Structural Diffusion Model and a Structure-Guided Refiner, leveraging a large-scale human-centric dataset for detailed, high-quality image synthesis.
Rui Zhao, Yuchao Gu, Jay Zhangjie Wu +5 authors
MotionDirector decouples motion adaptation from appearance in text-to-video diffusion models using a dual-path LoRAs architecture and appearance-debiased temporal loss.
Wang You, Wenshan Wu, Yaobo Liang +8 authors
The proposed EIPE-text framework iteratively extracts and refines narrative plans from a corpus to improve long-form text generation, resulting in more coherent and relevant narratives.
Bowen Pan, Rameswar Panda, SouYoung Jin +4 authors
Using natural language as a perceptual representation for vision-and-language navigation improves performance over visual feature-based approaches, especially with limited gold trajectories.
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