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。
Rui-Jie Zhu, Zixuan Wang, Kai Hua +30 authors
LoopLM, a family of pre-trained Looped Language Models, enhances reasoning by integrating iterative computation and entropy regularization during pre-training, achieving superior performance with better knowledge manipulation.
50 篇论文 · 按点赞排序
Yujia Zhang, Xiaoyang Wu, Yixing Lao +4 authors
Concerto, a minimalist model combining 3D self-distillation and 2D-3D joint embedding, achieves superior spatial feature learning and outperforms existing models in scene understanding and open-world perception.
Kimi Team, Yu Zhang, Zongyu Lin +57 authors
Kimi Linear, a hybrid linear attention architecture, outperforms full attention in various scenarios with improved efficiency and performance, using Kimi Delta Attention and Multi-Head Latent Attention.
Zhaoyang Yu, Jiayi Zhang, Huixue Su +10 authors
ReCode, a recursive code generation paradigm, unifies high-level planning and low-level action in a single representation, enhancing decision granularity and data efficiency in LLM-based agents.
Zhichao Wang, Dongyang Ma, Xinting Huang +6 authors
AutoDeco, a novel architecture, enables end-to-end generation by dynamically predicting decoding parameters, outperforming default strategies and achieving performance close to an oracle-tuned baseline.
Yufeng Cui, Honghao Chen, Haoge Deng +20 authors
Emu3.5, a large-scale multimodal world model, predicts next states in vision and language, enhanced with reinforcement learning and Discrete Diffusion Adaptation for efficient inference, achieving strong performance in various multimodal tasks.
Tongyi DeepResearch Team, Baixuan Li, Bo Zhang +53 authors
Tongyi DeepResearch, a large language model with agentic capabilities, achieves top performance in various deep research tasks through an end-to-end training framework and automatic data synthesis.
Xiaoxi Li, Wenxiang Jiao, Jiarui Jin +8 authors
DeepAgent, an end-to-end deep reasoning agent, autonomously performs thinking, tool discovery, and action execution using memory folding and reinforcement learning, outperforming baselines in various tool-use and application tasks.
Mingyi Deng, Lijun Huang, Yani Fan +22 authors
InteractComp evaluates search agents' ability to recognize and resolve query ambiguity through interaction, revealing significant gaps in current models' capabilities.
Qiushi Sun, Jingyang Gong, Yang Liu +6 authors
A unified multimodal code corpus and model enable high-quality code generation from both text and visual inputs, outperforming commercial models.
Shijian Wang, Jiarui Jin, Xingjian Wang +6 authors
Video-Thinker, a multimodal large language model, autonomously reasons with videos using intrinsic grounding and captioning capabilities, achieving state-of-the-art performance on various video reasoning benchmarks.
Rui Ye, Zhongwang Zhang, Kuan Li +12 authors
AgentFold, a novel proactive context management paradigm, enhances long-horizon task performance through dynamic context folding, achieving superior results on benchmarks compared to larger models and proprietary agents.
Yizhang Zhu, Liangwei Wang, Chenyu Yang +22 authors
A systematic taxonomy for data agents is introduced to clarify their autonomy levels and capabilities, addressing terminological ambiguity and guiding future research and development.
Chieh-Hsin Lai, Yang Song, Dongjun Kim +2 authors
Diffusion models are explored through variational, score-based, and flow-based perspectives, focusing on their mathematical foundations and applications in controllable generation and efficient sampling.
Siyin Wang, Jinlan Fu, Feihong Liu +11 authors
RoboOmni, a Perceiver-Thinker-Talker-Executor framework using end-to-end omni-modal LLMs, improves robotic manipulation by inferring user intentions from spoken dialogue, environmental sounds, and visual cues.
Guangting Zheng, Qinyu Zhao, Tao Yang +6 authors
FARMER, a unified generative framework combining Normalizing Flows and Autoregressive models, achieves competitive image synthesis performance with exact likelihoods and scalable training.
Zihao Wang, Xujing Li, Yining Ye +24 authors
Game-TARS, a generalist game agent trained with a unified action space, achieves superior performance across various domains and benchmarks through large-scale pre-training and efficient reasoning strategies.
Yuxuan Bian, Xin Chen, Zenan Li +4 authors
Video-As-Prompt (VAP) uses a reference video to guide a frozen Video Diffusion Transformer via a Mixture-of-Transformers expert, achieving state-of-the-art results in semantic-controlled video generation with strong zero-shot generalization.
Aayush Karan, Yilun Du
An iterative sampling algorithm enhances reasoning capabilities in base models without additional training, matching or outperforming reinforcement learning on single-shot tasks.
Yihe Deng, I-Hung Hsu, Jun Yan +7 authors
Supervised Reinforcement Learning (SRL) enhances small-scale LLMs' multi-step reasoning by generating internal monologues and using step-wise expert actions for richer learning signals, outperforming SFT and RLVR.
Junlong Li, Wenshuo Zhao, Jian Zhao +18 authors
Toolathlon is a benchmark for language agents that evaluates their performance across diverse, realistic, and complex multi-step workflows involving multiple software applications and tools.
Jingran Zhang, Ning Li, Justin Cui
ChatGPT Atlas excels in logical reasoning tasks within web games but struggles in real-time games requiring precise timing and motor control.
Haoge Deng, Ting Pan, Fan Zhang +8 authors
URSA, a discrete generative model, bridges the gap with continuous approaches in video generation by using iterative refinement, linearized metric paths, and resolution-dependent timestep shifting, achieving performance comparable to state-of-the-art continuous methods.
Taekhyun Park, Yongjae Lee, Hyerim Bae
PM4GRPO, a reasoning-aware Group Relative Policy Optimization, enhances policy models by incorporating process mining to align reasoning with a teacher model, outperforming existing methods.
Junyoung Seo, Rodrigo Mira, Alexandros Haliassos +6 authors
Lookahead Anchoring improves audio-driven human animation by using future keyframes as dynamic guides, enhancing lip synchronization, identity preservation, and visual quality.
Hao Li, Zhengyu Zou, Fangfu Liu +8 authors
InstanceGrounded Geometry Transformer (IGGT) unifies 3D reconstruction and instance-level understanding using a unified transformer and 3D-Consistent Contrastive Learning, supported by a new dataset InsScene-15K.
Sikuang Li, Chen Yang, Jiemin Fang +6 authors
WorldGrow, a hierarchical framework, generates large, continuous 3D environments with coherent geometry and realistic appearance using pre-trained 3D models and a coarse-to-fine generation strategy.
Inclusion AI, Bowen Ma, Cheng Zou +54 authors
Ming-Flash-Omni, an advanced Mixture-of-Experts model, enhances computational efficiency and multimodal intelligence, achieving state-of-the-art performance in speech recognition, image generation, and text-to-image generation.
Meituan LongCat Team, Xunliang Cai, Qilong Huang +8 authors
LongCat-Video, a 13.6B parameter video generation model based on the Diffusion Transformer framework, excels in efficient and high-quality long video generation across multiple tasks using unified architecture, coarse-to-fine generation, and block sparse attention.
Xiaoyu Liu, Chaoyou Fu, Chi Yan +15 authors
VITA-E, a dual-model embodied interaction framework, enables concurrent and interruptible vision-language-action capabilities, enhancing real-time user interaction.
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