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。
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.
50 篇论文 · 按点赞排序
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.
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.
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.
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.
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.
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.
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.
Yao Zhao, Rishabh Joshi, Tianqi Liu +3 authors
Sequence Likelihood Calibration (SLiC) is shown to be an effective and simpler alternative to Reinforcement Learning from Human Feedback (RLHF) for learning from human preferences in language models.
Yue Wang, Hung Le, Akhilesh Deepak Gotmare +3 authors
CodeT5+, a family of flexible encoder-decoder LLMs initialized with frozen LLMs and trained with a mixture of pretraining objectives, achieves state-of-the-art performance across various code-related tasks.
Karan Singhal, Tao Tu, Juraj Gottweis +28 authors
Med-PaLM 2, built on PaLM 2 and enhanced with medical fine-tuning and ensemble refinement, significantly improves physician-level performance in medical question answering by scoring 86.5% on MedQA and excelling on other clinical datasets.
James Betker
TorToise is an expressive multi-voice speech synthesis system that applies autoregressive transformers and DDPMs from image generation to text-to-speech.
Dong Zhang, Shimin Li, Xin Zhang +4 authors
SpeechGPT, a large language model with intrinsic cross-modal conversational abilities, demonstrates impressive performance in following multi-modal human instructions through a three-stage training strategy.
Sang Michael Xie, Hieu Pham, Xuanyi Dong +7 authors
Domain Reweighting with Minimax Optimization (DoReMi) uses a small proxy model to find optimal domain weights, improving language model performance and efficiency.
Canwen Xu, Yichong Xu, Shuohang Wang +3 authors
Super In-Context Learning integrates black-box LLMs with locally fine-tuned smaller models to enhance performance, multilinguality, and interpretability in supervised tasks.
Zhenhui Ye, Rongjie Huang, Yi Ren +5 authors
CLAPSpeech is a cross-modal contrastive pre-training framework that enhances prosody in text-to-speech systems by learning prosodic variance across contexts.
Wenhai Wang, Zhe Chen, Xiaokang Chen +8 authors
A unified LLM-based framework, VisionLLM, treats images as a foreign language, enabling open-ended vision tasks with language instructions and achieves performance comparable to detection-specific models.
Gaurav Verma, Ryan A. Rossi, Christopher Tensmeyer +2 authors
A method to detect visualness in text improves neural text-to-image generation by fine-tuning large vision-language models to classify visual and non-visual text accurately.
Can Qin, Shu Zhang, Ning Yu +10 authors
UniControl, a generative foundation model, integrates diverse visual controls and arbitrary language prompts, enabling precise image generation across multiple condition-to-image tasks while surpassing single-task control methods.
Jingye Chen, Yupan Huang, Tengchao Lv +3 authors
TextDiffuser, a two-stage model combining Transformer and diffusion models, generates visually appealing and coherent text images using text prompts and layouts, leveraging the MARIO-10M dataset for training and MARIO-Eval for evaluation.
Xuehai He, Weixi Feng, Tsu-Jui Fu +6 authors
A novel method, Discriminative Stable Diffusion (DSD), leverages pre-trained text-to-image diffusion models with attention-based prompt learning to achieve superior performance in few-shot image-text matching tasks.
Alexander H. Liu, Heng-Jui Chang, Michael Auli +2 authors
Self-distillation and online clustering enhance self-supervised speech representation learning by combining masked language modeling, contextualized embeddings, and discrete tokens, achieving superior performance in various downstream tasks.
Alexandros Lattas, Stylianos Moschoglou, Stylianos Ploumpis +3 authors
FitMe is a facial reflectance model with a differentiable rendering pipeline that generates high-fidelity human avatars from single or multiple images, offering accurate reflectance and identity preservation.
Ameya Prabhu, Zhipeng Cai, Puneet Dokania +3 authors
Research addresses online continual learning with focus on computational budgets, using a kNN classifier and pre-trained feature extractors to manage large datasets efficiently.
Chandan Singh, Aliyah R. Hsu, Richard Antonello +4 authors
SASC, a method for generating natural language explanations for black-box text modules, demonstrates reliability in synthetic, BERT model, and fMRI voxel contexts.
Tong Wu, Zhihao Fan, Xiao Liu +9 authors
Auto-Regressive Diffusion (AR-Diffusion) enhances text generation by dynamically varying denoising steps based on token position, improving efficiency and performance compared to existing diffusion models.
Samaneh Azadi, Akbar Shah, Thomas Hayes +2 authors
A text-conditioned human motion generation model, Make-An-Animation, achieves state-of-the-art performance by learning from large-scale image-text datasets and using a U-Net architecture for both spatial and temporal modeling.
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