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。
Wenyi Hong, Wenmeng Yu, Xiaotao Gu +74 authors
A vision-language model (VLM) named GLM-4.1V-Thinking, developed with a reasoning-centric training framework, achieves state-of-the-art performance across various tasks, including STEM problem solving, video understanding, and long document understanding, outperforming larger models on many benchmarks.
50 篇论文 · 按点赞排序
Kwai Keye Team, Biao Yang, Bin Wen +57 authors
Kwai Keye-VL, an 8-billion-parameter multimodal model, excels in short-video understanding and general vision-language tasks through a comprehensive pre-training and post-training process, including a five-mode data mixture and reinforcement learning.
Kuan Li, Zhongwang Zhang, Huifeng Yin +16 authors
WebSailor, a post-training methodology, enhances open-source LLMs with sophisticated reasoning to match proprietary systems in complex information-seeking tasks.
Zhaochen Su, Peng Xia, Hangyu Guo +12 authors
The survey outlines the evolution of multimodal AI from treating vision as static context to integrating it dynamically into the reasoning process, highlighting three stages and key challenges.
Maggie Huan, Yuetai Li, Tuney Zheng +6 authors
Reinforcement learning-tuned models generalize better across domains compared to supervised fine-tuned models in reasoning tasks, indicating a need to reconsider standard training methods.
Nan Chen, Mengqi Huang, Yihao Meng +1 authors
LongAnimation framework achieves long-term color consistency in animation colorization by integrating a dynamic global-local memory and color consistency reward.
Alexi Gladstone, Ganesh Nanduru, Md Mofijul Islam +7 authors
Energy-Based Transformers (EBTs) improve model performance and scalability across modalities by learning to verify predictions through unsupervised learning and energy minimization.
Jiacheng Chen, Ramin Mehran, Xuhui Jia +2 authors
A generative visual compositing framework using a diffusion model for scene editing and composition with source masking and simulated object jittering.
Guo-Hua Wang, Shanshan Zhao, Xinjie Zhang +9 authors
Ovis-U1, a 3-billion-parameter unified model, integrates multimodal understanding, text-to-image generation, and image editing using a diffusion-based visual decoder and bidirectional token refiner, achieving state-of-the-art performance across various benchmarks.
Chris Yuhao Liu, Liang Zeng, Yuzhen Xiao +9 authors
Skywork-Reward-V2, a suite of reward models trained on a large-scale, high-quality preference dataset, achieves state-of-the-art performance across various benchmarks by leveraging human-AI synergistic curation.
Fangfu Liu, Hao Li, Jiawei Chi +4 authors
LangScene-X uses a TriMap video diffusion model and Language Quantized Compressor to generate 3D consistent multi-modality information from sparse 2D views, enabling open-ended language queries and superior generalization.
Bo Liu, Leon Guertler, Simon Yu +9 authors
Self-play in zero-sum games using SPIRAL framework enhances reasoning capabilities in language models through continuous adaptation and transfer learning.
Boyuan Sun, Modi Jin, Bowen Yin +1 authors
DepthAnything-AC is a monocular depth estimation model that uses unsupervised consistency regularization and spatial distance constraints to handle complex environmental conditions and achieve zero-shot performance across various benchmarks.
Yilun Zhao, Kaiyan Zhang, Tiansheng Hu +15 authors
SciArena is a community-driven platform for evaluating foundation models on scientific literature tasks using collective intelligence and human voting.
Xingyang Li, Muyang Li, Tianle Cai +11 authors
Radial Attention, a scalable sparse attention mechanism, improves efficiency and preserves video quality in diffusion models by leveraging spatiotemporal energy decay.
Nina Konovalova, Maxim Nikolaev, Andrey Kuznetsov +1 authors
InnerControl enhances text-to-image diffusion models by enforcing spatial consistency across all diffusion steps using lightweight convolutional probes, improving control fidelity and generation quality.
Yifan Zhong, Fengshuo Bai, Shaofei Cai +11 authors
This survey categorizes and analyzes vision-language-action models through the lens of action tokenization, identifying strengths, limitations, and future directions.
Yue Ma, Qingyan Bai, Hao Ouyang +8 authors
Calligrapher, a diffusion-based framework, integrates advanced text customization with artistic typography using self-distillation, a localized style injection framework, and in-context generation to achieve high-quality, visually consistent typography.
Boyuan Sun, Jiaxing Zhao, Xihan Wei +1 authors
LLaVA-Scissor, a token compression strategy for video multimodal large language models, uses Semantic Connected Components to compress tokens effectively while maintaining semantic coverage and outperforming other methods.
Haonan Chen, Hong Liu, Yuping Luo +4 authors
MoCa, a two-stage framework, enhances pre-trained VLMs into effective bidirectional multimodal embedding models by addressing limitations through modality-aware continual pre-training and heterogeneous contrastive fine-tuning.
Jianzong Wu, Liang Hou, Haotian Yang +5 authors
VMoBA, a novel sparse attention mechanism, enhances Video Diffusion Models by improving efficiency and quality in generating long-duration, high-resolution videos.
The IntFold Team, Leon Qiao, Wayne Bai +4 authors
IntFold, a foundation model with a customized attention kernel, achieves accuracy comparable to AlphaFold3 and can predict various biomolecular structures and binding affinities using adapters and a confidence head for docking quality.
Shansan Gong, Ruixiang Zhang, Huangjie Zheng +4 authors
Diffusion large language models are applied to code generation, revealing their unique denoising processes and benefiting from a novel reinforcement learning sampling scheme.
Qiyue Gao, Xinyu Pi, Kevin Liu +21 authors
A benchmark framework evaluates the world modeling capabilities of Vision-Language Models, highlighting their limitations in perception and prediction.
Bowen Chen, Mengyi Zhao, Haomiao Sun +4 authors
XVerse enhances text-to-image generation by enabling precise and independent control over multiple subjects using token-specific text-stream modulation, improving image coherence and fidelity.
Jiajie Jin, Xiaoxi Li, Guanting Dong +5 authors
A hierarchical framework separates strategic planning from specialized execution in deep search tasks, improving answer quality and system efficiency over traditional RAG and agent-based systems.
Aurko Roy, Timothy Chou, Sai Surya Duvvuri +5 authors
The 2-simplicial Transformer improves token efficiency over standard Transformers, offering better performance on knowledge and reasoning tasks with a fixed token budget.
Rizwan Qureshi, Ranjan Sapkota, Abbas Shah +17 authors
The paper explores the development of Artificial General Intelligence by integrating insights from various fields, focusing on modular reasoning, memory, and multi-agent coordination, and highlights the challenges in achieving true intelligence.
Zhuoyang Zhang, Luke J. Huang, Chengyue Wu +4 authors
Locality-aware Parallel Decoding accelerates autoregressive image generation by enabling flexible parallelization and optimized generation ordering, significantly reducing latency without compromising quality.
Alexander Gambashidze, Li Pengyi, Matvey Skripkin +5 authors
A listener-augmented GRPO framework improves the accuracy and generalization of reward models for aligning text-to-image and text-to-video generative models with human preferences.
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