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。
Zhiyuan Liang, Dongwen Tang, Yuhao Zhou +11 authors
Drag-and-Drop LLMs generate task-specific parameters through prompt-conditioned parameter generation, achieving significant efficiency gains and cross-domain generalization without per-task training.
50 篇论文 · 按点赞排序
Hong Li, Houyuan Chen, Chongjie Ye +11 authors
Photometric stereo aims to recover high-quality surface normals under arbitrary lighting conditions, addressing challenges related to illumination-surface normal coupling and high-frequency geometric detail preservation.
Vishesh Tripathi, Tanmay Odapally, Indraneel Das +2 authors
A novel multimodal document chunking approach using Large Multimodal Models (LMMs) enhances RAG performance by accurately processing complex PDF documents, including multi-page tables and embedded visuals.
Guilherme Penedo, Hynek Kydlíček, Vinko Sabolčec +7 authors
A new pre-training dataset curation pipeline based on FineWeb supports multilingual LLMs by automatically adapting to any language, improving model performance and balancing dataset quality.
Chenyuan Wu, Pengfei Zheng, Ruiran Yan +19 authors
OmniGen2, a versatile generative model, introduces dual decoding pathways for text and images, preserves original text generation, and achieves competitive results with a new subject-driven benchmark.
Yifan Zhang, Chunli Peng, Boyang Wang +8 authors
Matrix-Game, a controllable game world generation model trained in a two-stage process, outperforms existing models by producing high-quality, action-controllable, and physically consistent Minecraft world videos.
Junying Chen, Zhenyang Cai, Pengcheng Chen +5 authors
ShareGPT-4o-Image and Janus-4o enable open research in photorealistic, instruction-aligned image generation through a large dataset and multimodal model.
Yunlong Lin, Zixu Lin, Kunjie Lin +8 authors
JarvisArt, an MLLM-driven agent, achieves superior photo retouching by understanding user intent and coordinating multiple retouching tools in Lightroom, outperforming GPT-4o on a novel benchmark.
Zehuan Huang, Haoran Feng, Yangtian Sun +3 authors
AnimaX creates multi-skeleton 3D animations by blending video diffusion model priors with skeleton-based control, using joint video-pose diffusion and shared positional encodings.
Tianchen Zhao, Ke Hong, Xinhao Yang +8 authors
PAROAttention reorganizes visual attention patterns to enable efficient sparsification and quantization, reducing memory and computational costs with minimal impact on performance.
Yuhao Wu, Yushi Bai, Zhiqiang Hu +2 authors
An incentivization-based reinforcement learning approach is used to develop a large language model capable of generating ultra-long, high-quality text without the need for synthetic data or supervised fine-tuning.
Jiaqi Li, Junshu Tang, Zhiyong Xu +6 authors
Hunyuan-GameCraft is a novel framework for high-dynamic interactive video generation in game environments that addresses limitations in dynamics, generality, and efficiency through unified input representation, hybrid history-conditioned training, and model distillation.
Liang Zeng, Yongcong Li, Yuzhen Xiao +8 authors
An automated data-curation pipeline for software engineering improves large language model performance on SWE tasks, achieving state-of-the-art results with and without test-time scaling techniques.
Boyu Gou, Zanming Huang, Yuting Ning +23 authors
Mind2Web 2 benchmark evaluates agentic search systems with a suite of realistic, long-horizon tasks, introducing an Agent-as-a-Judge framework to assess accuracy and source attribution.
Zengzhi Wang, Fan Zhou, Xuefeng Li +1 authors
Investigating mid-training strategies reveals that high-quality mathematical corpora and well-formatted chain-of-thought reasoning examples enhance reinforcement learning performance in language models, leading to the development of OctoThinker.
Jinming Wu, Zihao Deng, Wei Li +5 authors
MMSearch-R1, a reinforcement learning framework, enables large multimodal models to perform efficient, on-demand, multi-turn search in real-world environments, outperforming existing approaches.
Jungwoo Park, Taewhoo Lee, Chanwoong Yoon +2 authors
Outlier-Safe Pre-Training improves large language model quantization performance by preventing extreme activation outliers through innovative training techniques.
Zihan Wang, Rui Pan, Jiarui Yao +7 authors
Chain-of-Experts (CoE) improves performance and memory efficiency in mixture-of-experts models by iteratively routing tokens through experts within each layer.
Ilia Beletskii, Andrey Kuznetsov, Aibek Alanov
A new framework using consistency models enhances image inversion and editing efficiency, achieving top performance with fewer steps.
Advait Gupta, Rishie Raj, Dang Nguyen +1 authors
A neurosymbolic agent combines language models for fast subtask planning with A$^*$ search for detailed toolpaths, creating a cost-efficient multi-turn image editing solution.
Jun Cen, Chaohui Yu, Hangjie Yuan +9 authors
WorldVLA, an autoregressive action world model integrating vision-language-action (VLA) and world models, enhances performance through mutual understanding and generation, improving action prediction and sequence generation with an attention mask strategy.
Li Kang, Xiufeng Song, Heng Zhou +6 authors
VIKI-Bench and VIKI-R provide a benchmark and framework for evaluating and improving visual-driven cooperation among diverse embodied agents using vision-language models and reinforcement learning.
Polina Karpikova, Daniil Selikhanovych, Kirill Struminsky +3 authors
MADrive enhances scene reconstruction for autonomous driving by integrating visually similar 3D car assets from an external memory bank to achieve photorealistic synthesis of altered scenarios.
Tianyu Yu, Bo Ji, Shouli Wang +9 authors
RLPR, a verifier-free framework using LLM's token probability scores as reward signals, enhances reasoning capabilities across both general and mathematical domains, outperforming other methods in various benchmarks.
Jiaming Han, Hao Chen, Yang Zhao +6 authors
A multimodal framework uses a Text-Aligned Tokenizer (TA-Tok) to integrate vision and text into a unified space, employing a generative de-tokenizer with autoregressive and diffusion-based models for efficient and high-fidelity visual outputs.
He Zhu, Tianrui Qin, King Zhu +21 authors
Recently, Agentic AI has become an increasingly popular research field. However, we argue that current agent research practices lack standardization and scientific rigor, making it hard to conduct fair comparisons among methods. As a result, it is still unclear how different design choices in agent frameworks affect effectiveness, and measuring their progress remains challenging. In this work, we conduct a systematic empirical study on GAIA benchmark and BrowseComp to examine the impact of popular design choices in key agent components in a fair and rigorous manner. We find that the lack of a standard evaluation protocol makes previous works, even open-sourced ones, non-reproducible, with significant variance between random runs. Therefore, we introduce a more robust evaluation protocol to stabilize comparisons. Our study reveals which components and designs are crucial for effective agents, while others are redundant, despite seeming logical. Based on our findings, we build and open-source OAgents, a new foundation agent framework that achieves state-of-the-art performance among open-source projects. OAgents offers a modular design for various agent components, promoting future research in Agentic AI.
Zeqiang Lai, Yunfei Zhao, Haolin Liu +23 authors
Hunyuan3D 2.5, a suite of 3D diffusion models, advances shape and texture generation with a new LATTICE model and physical-based rendering in a multi-view architecture.
Michal Nazarczuk, Sibi Catley-Chandar, Thomas Tanay +3 authors
ViDAR uses diffusion-aware reconstruction to generate high-quality novel views of dynamic scenes from monocular video, outperforming existing methods in visual quality and geometric consistency.
Zhuowei Chen, Bingchuan Li, Tianxiang Ma +8 authors
A cross-pair dataset called Phantom-Data improves subject-to-video generation by enhancing prompt alignment and visual quality while maintaining identity consistency.
Jiaru Zou, Ling Yang, Jingwen Gu +4 authors
ReasonFlux-PRM, a novel trajectory-aware Process Reward Model, evaluates reasoning traces with step-level and trajectory-level supervision, enhancing performance in model distillation, reinforcement learning, and test-time scaling.
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