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。
Kuzma Khrabrov, Anton Ber, Artem Tsypin +10 authors
A new dataset and benchmark, $\nabla^2$DFT, enhances neural network potentials for molecular property prediction, Hamiltonian prediction, and conformational optimization with extensive, diverse molecular data.
50 篇论文 · 按点赞排序
Daixuan Cheng, Yuxian Gu, Shaohan Huang +3 authors
Instruction Pre-Training enhances language models by generating and incorporating instruction-response pairs into unsupervised multitask pre-training.
Alexander Nikulin, Ilya Zisman, Alexey Zemtsov +3 authors
XLand-100B is a large-scale dataset for in-context reinforcement learning, containing extensive learning histories in the XLand-MiniGrid environment.
Lital Binyamin, Yoad Tewel, Hilit Segev +3 authors
CountGen enhances text-to-image diffusion models by accurately controlling the number of depicted objects using features that carry object identity information, surpassing existing methods on benchmark datasets.
DeepSeek-AI, Qihao Zhu, Daya Guo +37 authors
DeepSeek-Coder-V2, a Mixture-of-Experts language model, excels in code-specific tasks by enhancing coding and mathematical reasoning capabilities while expanding language support and context length.
Denis Bobkov, Vadim Titov, Aibek Alanov +1 authors
StyleFeatureEditor combines W-space and F-space for effective image attribute manipulation in StyleGAN, balancing reconstruction quality and detailed editing.
Ziyu Liu, Tao Chu, Yuhang Zang +8 authors
A new benchmark (MMDU) and instruction tuning dataset (MMDU-45k) are introduced to evaluate and enhance large vision-language models' performance in multi-turn and multi-image conversations, highlighting the need for improved conversational instruction tuning data.
Jeffrey Li, Alex Fang, Georgios Smyrnis +56 authors
DataComp for Language Models (DCLM) provides a benchmark with a standardized corpus and recipes to improve language models through controlled dataset experiments, showing the importance of data curation for performance and compute efficiency.
Chufan Shi, Cheng Yang, Yaxin Liu +11 authors
ChartMimic is a new benchmark assessing large multimodal models' visually-grounded code generation for chart rendering, evaluating multi-level cognitive capabilities with diverse chart types and metrics.
Weiyun Wang, Shuibo Zhang, Yiming Ren +13 authors
NN-NIAH is a benchmark that evaluates MLLMs' comprehension of long multimodal documents across retrieval, counting, and reasoning tasks.
Ning-Hsu Wang, Yu-Lun Liu
A new depth estimation framework for 360-degree imagery uses pseudo-labels generated by perspective depth models to improve accuracy, especially in zero-shot scenarios.
Yuri Kuratov, Aydar Bulatov, Petr Anokhin +4 authors
BABILong benchmark evaluates large language models' reasoning across long documents, revealing that models use only a small portion of context and that recurrent memory transformers handle longer contexts better.
Changyu Chen, Zichen Liu, Chao Du +5 authors
A novel method using the implicit reward model from Direct Preference Optimization (DPO) to iteratively improve the alignment of large language models, achieving superior performance without external feedback.
Lingyun Zhang, Bin jin, Gaojian Ge +7 authors
A principle for leveraging human priors in data construction is proposed to improve small language models, demonstrating favorable performance on large benchmarks in resource-constrained settings.
Fei Wang, Wenxuan Zhou, James Y. Huang +4 authors
mDPO, a multimodal preference optimization method, addresses the unconditional preference problem and enhances model performance in multimodal large language models.
Byung-Kwan Lee, Sangyun Chung, Chae Won Kim +2 authors
The TroL model efficiently reduces the size of large language and vision models while maintaining or surpassing performance through a layer traversal technique that reuses layers token-wise.
Yuxuan Qiao, Haodong Duan, Xinyu Fang +6 authors
Prism is a framework that separates perception and reasoning in Vision Language Models, achieving competitive performance with smaller models and reduced costs.
Team GLM, Aohan Zeng, Bin Xu +53 authors
GLM-4 language models, trained on diverse datasets, outperform GPT-4 in various tasks and feature tool integration capabilities that match or surpass those of GPT-4 All Tools.
Seo Hyun Kim, Kai Tzu-iunn Ong, Taeyoon Kwon +7 authors
Theanine enhances LLM response generation by integrating memory timelines, providing contextual cues for better understanding and recall of past events in dialogues.
Xinyu Fang, Kangrui Mao, Haodong Duan +4 authors
MMBench-Video assesses large vision-language models' video understanding capabilities through a quantitative benchmark using YouTube videos and human-annotated questions.
Hengyi Wang, Haizhou Shi, Shiwei Tan +6 authors
The MultiModal Needle-in-a-haystack benchmark assesses the long-context capabilities of Multimodal Large Language Models by evaluating their performance in locating sub-images within a set of images based on textual instructions.
Holy Lovenia, Rahmad Mahendra, Salsabil Maulana Akbar +58 authors
SEACrowd consolidates a comprehensive resource hub and benchmarks for evaluating AI models across nearly 1,000 Southeast Asian languages to address the lack of representation and improve cultural sensitivity in AI.
Qingyun Li, Zhe Chen, Weiyun Wang +37 authors
OmniCorpus, a large-scale image-text interleaved dataset, enhances multimodal model research by increasing volume, diversity, and flexibility over existing datasets.
Xubing Ye, Yukang Gan, Xiaoke Huang +3 authors
VoCo-LLaMA, a vision token compression method using LLMs, enhances computational efficiency and maintains performance in VLMs, facilitating scalable multi-modal applications.
Hoyeon Chang, Jinho Park, Seonghyeon Ye +4 authors
LLMs acquire factual knowledge during pretraining through a process involving memorization, generalization, and forgetting, which is influenced by factors like data volume, batch size, and data duplication.
Yiwen Chen, Tong He, Di Huang +9 authors
MeshAnything, integrating VQ-VAE and a shape-conditioned transformer, generates high-quality, artist-created meshes with significantly reduced face count, improving efficiency compared to existing mesh extraction methods.
Hasan Abed Al Kader Hammoud, Umberto Michieli, Fabio Pizzati +4 authors
Proposed method improves model alignment during the merging of large language models by incorporating synthetic safety and domain-specific data.
Sachit Menon, Richard Zemel, Carl Vondrick
A novel method, whiteboard-of-thought prompting, enhances the visual reasoning capabilities of large language models by leveraging their ability to generate images using coding libraries, achieving superior accuracy on spatial and visual reasoning tasks compared to traditional chain-of-thought methods.
Nikita Starodubcev, Mikhail Khoroshikh, Artem Babenko +1 authors
iCD, a consistency distillation framework, enhances text-to-image diffusion models for both high-quality synthesis and image inversion, enabling zero-shot text-guided image editing.
Minh Huynh Nguyen, Thang Phan Chau, Phong X. Nguyen +1 authors
AgileCoder, a multi-agent system integrating Agile Methodology, enhances software development efficiency through sprints and dynamic code graph generation.
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