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。
Bang Liu, Xinfeng Li, Jiayi Zhang +44 authors
This survey covers the design, evaluation, and improvement of intelligent agents based on modular, brain-inspired architectures, focusing on self-enhancement, multi-agent collaboration, and safety in AI systems.
50 篇论文 · 按点赞排序
Cong Wei, Bo Sun, Haoyu Ma +10 authors
MoCha generates realistic talking character animations from speech and text using a speech-video attention mechanism and joint training on speech-labeled and text-labeled data, enabling multi-character conversations and superior realism.
Siyuan Li, Luyuan Zhang, Zedong Wang +8 authors
MergeVQ and MergeAR enhance MIM with VQ by integrating token merging and KV cache compression, improving both image generation and visual representation learning.
Nikai Du, Zhennan Chen, Zhizhou Chen +5 authors
TextCrafter is a progressive method that enhances the generation of intricate visual text within images, addressing issues like distortion, blurriness, and omissions through token focus enhancement.
Abhay Kumar, Louis Owen, Nilabhra Roy Chowdhury +1 authors
ZClip is an adaptive gradient clipping algorithm that uses z-score-based anomaly detection to dynamically adjust clipping thresholds and prevent large gradient spikes during LLM training.
Itay Nakash, Nitay Calderon, Eyal Ben David +2 authors
AdaptiVocab, a novel vocabulary adaptation method, enhances the efficiency of Large Language Models in domain-specific applications without sacrificing performance by reducing token usage.
Shengqiong Wu, Weicai Ye, Jiahao Wang +8 authors
Any2Caption framework improves video generation quality and controllability by decoupling condition interpretation and leveraging multimodal large language models.
Xiangyu Zhao, Peiyuan Zhang, Kexian Tang +7 authors
RISEBench evaluates Reasoning-Informed Visual Editing by assessing temporal, causal, spatial, and logical reasoning in multimodal models like GPT-4o-Native.
Yuxuan Luo, Zhengkun Rong, Lizhen Wang +3 authors
A diffusion transformer framework, DreamActor-M1, enhances image-based human animation through hybrid guidance, achieving robust control, multi-scale adaptability, and long-term temporal coherence.
Zhenyi Liao, Qingsong Xie, Yanhao Zhang +4 authors
The study enhances visual-spatial reasoning in multi-modal large language models through GRPO training using the VSI-100k dataset, demonstrating significant performance improvements over base models.
Junhao Cheng, Yuying Ge, Yixiao Ge +2 authors
AnimeGamer uses Multimodal Large Language Models (MLLMs) and a video diffusion model to generate contextually consistent, dynamic, and high-quality video clips for an infinite game simulation.
Jingcheng Hu, Yinmin Zhang, Qi Han +3 authors
Open-Reasoner-Zero achieves superior performance on reasoning benchmarks using a minimalist PPO approach, requiring fewer training steps than DeepSeek-R1-Zero.
Zichen Liu, Changyu Chen, Wenjun Li +5 authors
A study of R1-Zero training identifies pretraining effects on RL performance and proffers Dr. GRPO to enhance token efficiency, achieving superior accuracy on AIME 2024.
Zijun Liu, Peiyi Wang, Runxin Xu +5 authors
Self-Principled Critique Tuning enhances pointwise generative reward modeling for large language models, improving scalability and quality compared to existing methods.
Nuo Chen, Zhiyuan Hu, Qingyun Zou +4 authors
JudgeLRM models trained with reinforcement learning outperform existing models, especially in tasks requiring deep reasoning.
Zhiyuan Yan, Junyan Ye, Weijia Li +7 authors
GPT-4o demonstrates strong image generation and editing capabilities with a proposed auto-regressive and diffusion-based architecture, surpassing existing methods and highlighting specific limitations and safety concerns.
Olga Golovneva, Tianlu Wang, Jason Weston +1 authors
A new attention mechanism, Multi-Token Attention (MTA), enhances LLM performance by conditioning attention weights on multiple query and key vectors, improving context search and standard language modeling tasks.
Kai Liu, Wei Li, Lai Chen +8 authors
JavisDiT, a Joint Audio-Video Diffusion Transformer, generates high-quality synchronized audio-video content using a Hierarchical Spatial-Temporal Synchronized Prior Estimator, excelling in a new benchmark with a robust synchronization metric.
Qiyuan Zhang, Fuyuan Lyu, Zexu Sun +7 authors
A survey of test-time scaling for large language models, structured into dimensions of what, how, where, and how well to scale, with an emphasis on its applications and future challenges.
Fa-Ting Hong, Zunnan Xu, Zixiang Zhou +5 authors
ACTalker is an end-to-end video diffusion framework that uses a parallel mamba structure to generate talking head videos with flexible, multi-modal control, ensuring natural coordination across temporal and spatial dimensions.
Yue Liu, Jiaying Wu, Yufei He +6 authors
This survey reviews efficient inference methods for Large Reasoning Models to reduce token usage and memory consumption while maintaining reasoning quality, categorizing methods into explicit compact Chain-of-Thought and implicit latent CoT.
Wei Shen, Guanlin Liu, Zheng Wu +5 authors
This study enhances RLHF performance by addressing reward hacking and response diversity through a hybrid reward system and a novel prompt-selection method.
Xiaoye Qu, Yafu Li, Zhaochen Su +15 authors
The survey focuses on addressing inefficiencies in reasoning efficiency of Large Reasoning Models (LRMs) by examining methods across their lifecycle and identifying common inefficiency patterns.
Yubo Wang, Xueguang Ma, Ping Nie +7 authors
ScholarCopilot is a unified framework that enhances large language models by dynamically retrieving and integrating scholarly citations, improving accuracy and effectiveness in academic text generation.
Hanyang Wang, Fangfu Liu, Jiawei Chi +1 authors
VideoScene distills video diffusion models for efficient one-step 3D scene generation from sparse views, enhancing speed and accuracy compared to previous methods.
Liang Pan, Zeshi Yang, Zhiyang Dou +5 authors
A unified transformer-based policy, TokenHSI, improves versatility in Human-Scene Interactions by integrating multiple interaction skills through shared and task-specific tokens.
Xindi Yang, Baolu Li, Yiming Zhang +8 authors
A two-stage framework combines a Vision Language Model for coarse motion planning with a Video Diffusion Model to generate physically plausible videos.
Zhengcong Fei, Debang Li, Di Qiu +8 authors
SkyReels-A2, an open-source framework, generates high-quality, element-controlled videos from textual prompts using a novel image-text embedding model, optimized inference pipeline, and A2 Bench for systematic evaluation.
Xiaomin Yu, Pengxiang Ding, Wenjie Zhang +7 authors
A three-stage framework synthesizes high-quality multimodal datasets using text to generate captions and transform them into visual representations, enabling cost-effective training of vision-language models.
Weizhi Wang, Yu Tian, Linjie Yang +2 authors
Open-Qwen2VL is a 2B-parameter multimodal LLM pre-trained efficiently on image-text pairs with improved data filtering and sequence packing, outperforming state-of-the-art models on various benchmarks.
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