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。
Tim Dettmers, Artidoro Pagnoni, Ari Holtzman +1 authors
QLoRA enables efficient finetuning of large language models using 4-bit quantization and Low Rank Adapters, achieving high performance with reduced memory usage.
50 篇论文 · 按点赞排序
Ronen Eldan, Yuanzhi Li
TinyStories, a synthetic dataset of simplified stories, enables small or simple language models to produce fluent and coherent text with reasoning capabilities using an evaluation framework based on human-like grading.
Xingang Pan, Ayush Tewari, Thomas Leimkühler +3 authors
DragGAN enables precise user-controlled manipulation of image content through interactive point dragging, improving realism and flexibility over existing GAN-based methods.
Raymond Li, Loubna Ben Allal, Yangtian Zi +64 authors
StarCoder, a 15.5B parameter LLM trained on 1 trillion tokens, outperforms other open Code LLMs across multiple languages and fine-tuned Python, with safety enhancements and publicly available under the Open Responsible AI Model license.
Chunting Zhou, Pengfei Liu, Puxin Xu +12 authors
A 65B parameter LLaMa language model trained with minimal instructional data matches or outperforms models with extensive human preference modeling in most cases, indicating that pretraining is predominantly responsible for knowledge acquisition.
Bo Peng, Eric Alcaide, Quentin Anthony +27 authors
A new architecture, RWKV, combines the parallelizable training of Transformers with the efficient inference of RNNs, achieving linear scaling and matching performance.
Shunyu Yao, Dian Yu, Jeffrey Zhao +4 authors
The Tree of Thoughts framework enhances language models' problem-solving capabilities by enabling the exploration and evaluation of multiple reasoning paths and thoughts.
Zalán Borsos, Matt Sharifi, Damien Vincent +3 authors
SoundStorm, a non-autoregressive audio generation model, delivers high-quality and consistent audio two orders of magnitude faster than autoregressive methods.
Shanchuan Lin, Bingchen Liu, Jiashi Li +1 authors
Proposed solutions address flaws in diffusion noise schedules and samplers to ensure consistency between training and inference, enabling more accurate image generation.
Guanzhi Wang, Yuqi Xie, Yunfan Jiang +5 authors
Voyager is an LLM-powered agent in Minecraft that autonomously explores, learns, and discovers skills through a curriculum, skill library, and prompting mechanism, demonstrating superior lifelong learning and task-solving capabilities.
Gabriela Ben Melech Stan, Diana Wofk, Scottie Fox +8 authors
Latent Diffusion Model for 3D (LDM3D) generates RGBD images from text prompts and is used to create immersive 360-degree-view experiences.
Jinyang Li, Binyuan Hui, Ge Qu +13 authors
Bird, a large-scale text-to-SQL benchmark, addresses challenges posed by big databases, emphasizing database values, external knowledge, and SQL efficiency.
Youngjin Jin, Eugene Jang, Jian Cui +3 authors
DarkBERT, a language model trained on Dark Web data, outperforms existing models in various use cases due to its specialized training on the Dark Web's unique linguistic characteristics.
Zhengyi Wang, Cheng Lu, Yikai Wang +4 authors
Variational score distillation addresses issues in text-to-3D generation by modeling 3D parameters as random variables, improving sample quality and diversity.
Vipul Raheja, Dhruv Kumar, Ryan Koo +1 authors
CoEdIT, a fine-tuned large language model, enhances text editing quality and efficiency with diverse instructions, achieving state-of-the-art performance and user preference.
Lili Yu, Dániel Simig, Colin Flaherty +3 authors
Megabyte, a multi-scale decoder architecture, enables efficient byte-level modeling of long sequences through sub-quadratic self-attention, larger feedforward layers, and improved parallelism, achieving performance competitive with subword models and state-of-the-art results.
Renrui Zhang, Zhengkai Jiang, Ziyu Guo +5 authors
A training-free and fine-tuning variant of the Segment Anything Model (SAM), PerSAM and PerSAM-F, achieves personalized image and video segmentation using a single reference image and minimal fine-tuning, improving performance on personalized and dreambooth applications.
Yuxin Wen, John Kirchenbauer, Jonas Geiping +1 authors
A novel technique called Tree-Ring Watermarking robustly fingerprints diffusion model outputs by embedding a pattern into the initial noise vector and is invariant to various image transformations.
Nouha Dziri, Ximing Lu, Melanie Sclar +13 authors
Transformers solve compositional tasks by reducing multi-step reasoning to linearized subgraph matching, leading to performance decay with increased complexity.
Zeyue Xue, Guanglu Song, Qiushan Guo +4 authors
A text-conditional image diffusion model named RAPHAEL generates high-quality and artistic images by stacking mixture-of-experts layers, outperforming existing models in style switching and zero-shot FID score.
Zhibin Gou, Zhihong Shao, Yeyun Gong +4 authors
CRITIC, a framework that uses external tools to validate and improve LLM outputs, consistently enhances performance across various tasks.
Rohan Anil, Andrew M. Dai, Orhan Firat +125 authors
PaLM 2, a Transformer-based language model, improves multilingual and reasoning capabilities with enhanced efficiency and performance across various tasks compared to its predecessor.
Shashank Rajput, Nikhil Mehta, Anima Singh +10 authors
A generative retrieval model using Semantic IDs and a Transformer sequence-to-sequence approach improves recommendation system performance and generalization, especially for cold-start items.
Jing Yu Koh, Daniel Fried, Ruslan Salakhutdinov
A method to integrate text-only large language models with image encoders and decoders through embedding space mapping achieves multimodal capabilities including image retrieval, generation, and dialogue.
Omri Avrahami, Kfir Aberman, Ohad Fried +2 authors
The method decomposes a single image into distinct text tokens for each concept and fine-tunes textual embeddings using masked diffusion loss to enable synthesis with fine-grained control.
Ning Ding, Yulin Chen, Bokai Xu +6 authors
A large-scale dataset, UltraChat, and a fine-tuned model, UltraLLaMA, are created to improve open-source chat language models' performance.
Yabo Zhang, Yuxiang Wei, Dongsheng Jiang +3 authors
ControlVideo, a training-free framework adapted from ControlNet, improves text-to-video generation by ensuring appearance coherence, reducing flicker, and efficiently producing long videos.
Joseph Zhu, Peiye Zhuang
Reformulating optimization loss with diffusion prior and introducing depth supervision improve text-to-3D synthesis quality and multi-view consistency.
Ilia Shumailov, Zakhar Shumaylov, Yiren Zhao +3 authors
Model dementia, caused by the use of model-generated content in training, affects the quality of generative models like LLMs, VAEs, and GMMs, highlighting the need for genuine human interaction data.
Minghua Liu, Ruoxi Shi, Kaiming Kuang +6 authors
OpenShape learns multi-modal representations for text, images, and 3D point clouds, achieving superior zero-shot performance and fine-grained interactions through novel strategies and hard negative mining.
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