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。
50 篇论文 · 按点赞排序
Weijia Li, Jun He, Junyan Ye +5 authors
CrossViewDiff, a cross-view diffusion model, enhances satellite-to-street view synthesis by incorporating structural and textural controls, achieving superior results across various datasets.
Benjamin Cohen-Wang, Harshay Shah, Kristian Georgiev +1 authors
ContextCite is a method for attributing parts of the context used by language models to generate statements, aiding in verification, improving response quality, and detecting poisoning attacks.
Liuhan Chen, Zongjian Li, Bin Lin +6 authors
A novel omni-dimensional VAE (OD-VAE) is proposed to improve video compression by integrating both spatial and temporal dimensions, enhancing the efficiency and quality of latent video diffusion models.
Atnafu Lambebo Tonja, Bonaventure F. P. Dossou, Jessica Ojo +8 authors
InkubaLM, a small language model with 0.4 billion parameters, outperforms larger models in various tasks including sentiment analysis and demonstrates remarkable consistency across multiple languages, challenging the necessity for substantial resources in effective language models.
Xi Chen, Haosen Yang, Sheng Jin +2 authors
FrozenSeg combines spatial knowledge from a localization model and semantic knowledge from a vision-language model to enhance mask proposal quality in open-vocabulary segmentation.
Blair Yang, Fuyang Cui, Keiran Paster +4 authors
Report cards offer a human-interpretable evaluation framework for large language models, addressing limitations of traditional benchmarks.
Michael Burnham, Kayla Kahn, Ryan Yank Wang +1 authors
Political DEBATE models achieve high efficiency and accuracy in zero-shot and few-shot political document classification using minimal training data and are openly sourced.
Vage Egiazarian, Denis Kuznedelev, Anton Voronov +5 authors
Vector quantization techniques improve the compression rate of large text-to-image diffusion models without compromising their image quality and textual alignment.
Juncan Deng, Shuaiting Li, Zeyu Wang +3 authors
VQ4DiT, a post-training vector quantization method, enhances DiT models for efficient inference on edge devices by calibrating both codebooks and assignments.
Takuhiro Kaneko, Hirokazu Kameoka, Kou Tanaka +1 authors
FastVoiceGrad is a one-step diffusion-based voice conversion technique that matches or exceeds the performance of multi-step diffusion models while significantly improving inference speed.
Rohan Jha, Bo Wang, Michael Günther +3 authors
Jina-ColBERT-v2 enhances ColBERT's performance and reduces storage requirements through improvements to its architecture and training pipeline, benefiting multilingual retrieval tasks.
Jaesung Huh, Joon Son Chung, Arsha Nagrani +4 authors
A review of the VoxCeleb Speaker Recognition Challenges (VoxSRC) from 2019 to 2023, detailing performance trends, methodological advancements, and future research directions in speaker verification and diarisation.
Xiangyuan Xue, Zeyu Lu, Di Huang +2 authors
GenAgent, an LLM-based framework, generates complex workflows automatically, offering flexibility and scalability, and outperforms baselines in various task evaluations.
Leandro Carísio Fernandes, Gustavo Bartz Guedes, Thiago Soares Laitz +4 authors
A new dataset and two pipelines for summarizing scientific articles into survey sections are introduced and evaluated, emphasizing the impact of retrieval quality and configuration.
M. Maruf, Arka Daw, Kazi Sajeed Mehrab +19 authors
Pre-trained vision-language models are evaluated for their effectiveness in answering biologically relevant questions using a novel dataset without additional fine-tuning, exploring prompting techniques and reasoning hallucination.
Sungduk Yu, Brian L. White, Anahita Bhiwandiwalla +4 authors
ClimDetect uses vision transformers to analyze climate snapshots, offering a standardized dataset for climate change detection and attribution.
Muhammad Fawi
CURLoRA, a method that uses CUR matrix decomposition in Low-Rank Adaptation (LoRA), effectively mitigates catastrophic forgetting and reduces trainable parameters in fine-tuning large language models.
Abu Adnan Sadi, Mohammad Ashrafuzzaman Khan, Lubaba Binte Saber
A transformer-based system for differential diagnosis in healthcare achieves high accuracy through data processing, diversification, and multi-label classification.
Qihua Chen, Yue Ma, Hongfa Wang +7 authors
A diffusion-based method called Follow-Your-Canvas addresses low-quality content generation and GPU memory limitations in video outpainting by distributing the task across spatial windows and considering relative positional relation with the source video.
Sully F. Chen, Robert J. Steele, Beakal Lemeneh +2 authors
A multi-omic nucleotide-peptide foundation model learns joint representations of biosequences, captures structural information without prior structural training, and achieves state-of-the-art results in predicting peptide-nucleotide interactions and mutations impact.
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