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。
Binyuan Hui, Jian Yang, Zeyu Cui +14 authors
Qwen2.5-Coder series demonstrates state-of-the-art code generation, completion, reasoning, and repair capabilities using the Qwen2.5 architecture with over 5.5 trillion tokens of training data.
50 篇论文 · 按点赞排序
Aviral Kumar, Vincent Zhuang, Rishabh Agarwal +15 authors
The multi-turn online reinforcement learning approach SCoRe improves LLM self-correction using self-generated data and regularization, achieving state-of-the-art performance.
Matt Deitke, Christopher Clark, Sangho Lee +48 authors
Molmo, a new family of open-weight VLMs, excels through a high-quality human-annotated image caption dataset and a diverse fine-tuning dataset, outperforming both proprietary and open models in benchmark and human evaluations.
Shitao Xiao, Yueze Wang, Junjie Zhou +6 authors
OmniGen is a unified diffusion model for image generation that supports diverse tasks without additional modules, emphasizing simplicity, knowledge transfer, and reasoning capabilities.
Xinlong Wang, Xiaosong Zhang, Zhengxiong Luo +22 authors
Emu3, a transformer-based multimodal model trained exclusively with next-token prediction, outperforms existing diffusion and compositional models in generation and perception tasks.
Jianwen Jiang, Chao Liang, Jiaqi Yang +3 authors
An audio-only conditioned video diffusion model, Loopy, improves natural and high-quality human video generation without auxiliary spatial signals.
Vadim Titov, Madina Khalmatova, Alexandra Ivanova +2 authors
A modified diffusion sampling process using self-guidance and noise rescaling achieves high-quality image editing with preserved structure and appearance without requiring diffusion model fine-tuning.
Zifan Zheng, Yezhaohui Wang, Yuxin Huang +4 authors
This survey explores the interpretability and underlying mechanisms of attention heads within Large Language Models to understand their reasoning processes.
Haoran Wei, Chenglong Liu, Jinyue Chen +9 authors
GOT, a unified OCR model with 580M parameters, handles various types of characters and OCR tasks and offers advanced features like region-level recognition and dynamic resolution.
Niklas Muennighoff, Luca Soldaini, Dirk Groeneveld +21 authors
A sparse Mixture-of-Experts language model with 7 billion parameters achieves superior performance by using only 1 billion parameters per input token and outperforms larger models in various experiments.
Peng Wang, Shuai Bai, Sinan Tan +16 authors
The Qwen2-VL Series uses Naive Dynamic Resolution and Multimodal Rotary Position Embedding to enhance visual processing and achieves competitive performance on multimodal benchmarks.
Wenliang Dai, Nayeon Lee, Boxin Wang +7 authors
NVML 1.0, a family of multimodal large language models, achieves state-of-the-art results in vision-language tasks by combining text-only and multimodal training, utilizing a new architecture and dataset strategy.
Bofei Gao, Feifan Song, Yibo Miao +21 authors
Large Language Models (LLMs) exhibit remarkably powerful capabilities. One of the crucial factors to achieve success is aligning the LLM's output with human preferences. This alignment process often requires only a small amount of data to efficiently enhance the LLM's performance. While effective, research in this area spans multiple domains, and the methods involved are relatively complex to understand. The relationships between different methods have been under-explored, limiting the development of the preference alignment. In light of this, we break down the existing popular alignment strategies into different components and provide a unified framework to study the current alignment strategies, thereby establishing connections among them. In this survey, we decompose all the strategies in preference learning into four components: model, data, feedback, and algorithm. This unified view offers an in-depth understanding of existing alignment algorithms and also opens up possibilities to synergize the strengths of different strategies. Furthermore, we present detailed working examples of prevalent existing algorithms to facilitate a comprehensive understanding for the readers. Finally, based on our unified perspective, we explore the challenges and future research directions for aligning large language models with human preferences.
Sushant Gautam, Andrea Storås, Cise Midoglu +4 authors
Kvasir-VQA is a dataset with question-and-answer annotations for GI diagnostics, supporting image captioning, VQA, synthetic image generation, object detection, and classification.
Zecheng He, Bo Sun, Felix Juefei-Xu +14 authors
Imagine yourself is a tuning-free diffusion model for personalized image generation that enhances identity preservation, text alignment, and visual quality through synthetic data generation, parallel attention architecture, and multi-stage fine-tuning.
Ilya Gusev
A benchmark evaluates language models' role-playing abilities using automated components to simulate and judge dialogues.
Liqiang Jing, Zhehui Huang, Xiaoyang Wang +6 authors
DSBench evaluates large language and vision-language models on realistic data science tasks, revealing significant performance gaps that indicate the need for improved autonomous agents.
Fan Zhou, Zengzhi Wang, Qian Liu +2 authors
ProX, a novel framework treating data refinement as a programming task, enhances large language model pre-training by generating fine-grained operations for each example, outperforming traditional human-crafted rule-based methods significantly across various benchmarks and domains.
Qingkai Fang, Shoutao Guo, Yan Zhou +3 authors
LLaMA-Omni integrates a speech encoder, adaptor, LLM, and decoder for low-latency and high-quality speech interaction with LLMs, outperforming previous models in both content and style.
Praveen K Kanithi, Clément Christophe, Marco AF Pimentel +7 authors
MEDIC framework evaluates Large Language Models across five clinical dimensions to guide model selection in healthcare applications, identifying performance trade-offs and ensuring practical implementation.
Ye Bai, Haonan Chen, Jitong Chen +35 authors
Seed-Music generates high-quality music with style control using a unified framework combining auto-regressive language modeling and diffusion approaches, supporting vocal music generation and post-production editing.
Xidong Wang, Dingjie Song, Shunian Chen +2 authors
LongLLaVA, a hybrid MLLM combining Mamba and Transformer blocks, enhances video and image understanding with improved efficiency and throughput.
Zekun Wang, King Zhu, Chunpu Xu +14 authors
MIO, a novel foundation model, achieves competitive performance in multimodal tasks through an end-to-end, autoregressive approach using causal multimodal modeling.
Zhifei Xie, Changqiao Wu
Mini-Omni is an end-to-end audio-based conversational model enabling real-time speech interaction and efficient speech generation.
Abhilash Nandy, Yash Agarwal, Ashish Patwa +5 authors
Current Vision-Language models struggle with satirical image detection, understanding, and completion tasks on a newly released high-quality dataset called YesBut.
Xiaotian Han, Yiren Jian, Xuefeng Hu +8 authors
InfiMM-WebMath-40B, a large multimodal dataset, significantly improves mathematical reasoning capabilities in LLMs, achieving state-of-the-art performance on multimodal benchmarks.
Run Luo, Haonan Zhang, Longze Chen +13 authors
MMEvol, a multimodal instruction data evolution framework, enhances the capabilities of Multimodal Large Language Models by generating diverse and complex image-text instruction datasets, leading to improved performance across various vision-language tasks.
jiajie Zhang, Yushi Bai, Xin Lv +8 authors
A new pipeline and dataset enable long-context LLMs to generate responses with precise sentence-level citations, improving their trustworthiness and outperforming existing models.
Johannes Schmude, Sujit Roy, Will Trojak +26 authors
Prithvi WxC, a 2.3 billion parameter foundation model using an encoder-decoder architecture with transformer concepts, addresses weather forecasting, downscaling, and extreme events estimation with a mixed masked reconstruction and forecasting objective.
Rogerio Bonatti, Dan Zhao, Francesco Bonacci +8 authors
Windows Agent Arena provides a general, reproducible environment for evaluating multi-modal agent performance on Windows OS tasks, with scalability and parallelizability features.
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