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。
Weikai Huang, Jieyu Zhang, Sijun Li +14 authors
A unified 3D object detection framework with a large-scale dataset enables open-world detection with multiple prompt types and geometric cue integration.
50 篇论文 · 按点赞排序
Team Seedance, De Chen, Liyang Chen +168 authors
Seedance 2.0 is a multi-modal audio-video generation model that supports text, image, audio, and video inputs with improved generation quality and speed.
Yang Liu, Enxi Wang, Yufei Gao +6 authors
MEDS is a memory-enhanced dynamic reward shaping framework that improves sampling diversity in reinforcement learning for large language models by identifying and penalizing recurrent error patterns through clustering of historical behavioral signals.
Fei Tang, Zhiqiong Lu, Boxuan Zhang +4 authors
ClawGUI presents an open-source framework that addresses key challenges in GUI agent development through unified reinforcement learning, standardized evaluation, and cross-platform deployment capabilities.
Team HY-World, Chenjie Cao, Xuhui Zuo +42 authors
HY-World 2.0 is a multi-modal world model framework that generates high-fidelity 3D Gaussian Splatting scenes from diverse inputs using specialized modules for panorama generation, trajectory planning, world expansion, and composition, along with an enhanced rendering platform for interactive 3D exploration.
Ali Slim, Haydar Hamieh, Jawad Kotaich +5 authors
QuanBench+ evaluates large language models on quantum code generation across multiple frameworks using functional testing and repair-based feedback, revealing significant progress but persistent dependence on framework-specific knowledge.
Mingyu Ouyang, Siyuan Hu, Kevin Qinghong Lin +2 authors
GameWorld presents a standardized benchmark for evaluating multimodal large language model agents in video games, featuring diverse games and verified metrics for comprehensive assessment.
Yaxuan Li, Yuxin Zuo, Bingxiang He +8 authors
On-policy distillation dynamics in large language models depend on compatible thinking patterns between teacher and student models, with successful distillation characterized by alignment on high-probability tokens and requiring teachers to provide novel capabilities beyond student training data.
Haozhe Wang, Cong Wei, Weiming Ren +3 authors
Training reward models to generate multi-dimensional critiques improves visual generation through both enhanced reinforcement learning rewards and test-time refinement loops, achieving state-of-the-art performance with reduced training data requirements.
Linhao Yu, Tianmeng Yang, Siyu Ding +8 authors
KnowRL is a knowledge-guided reinforcement learning framework that improves reasoning in language models by optimizing compact, interaction-aware guidance subsets through constrained subset search and addressing pruning interaction paradoxes.
Xiangru Jian, Hao Xu, Wei Pang +13 authors
FORGE introduces a high-quality multimodal manufacturing dataset with fine-grained domain semantics to evaluate MLLMs on real-world tasks, revealing that domain-specific knowledge rather than visual grounding limits performance, and demonstrating that supervised fine-tuning on structured annotations significantly improves accuracy.
Zunhai Su, Hengyuan Zhang, Wei Wu +17 authors
Transformers face challenges from Attention Sink phenomenon where excessive attention focuses on uninformative tokens, impacting interpretability and performance, necessitating comprehensive research survey addressing fundamental usage, mechanistic understanding, and strategic mitigation approaches.
Eunbi Choi, Kibong Choi, Sehyun Chun +55 authors
EXAONE 4.5 is an open-weight vision language model that integrates a visual encoder into EXAONE 4.0, achieving enhanced document understanding and general language capabilities through targeted data curation and extended context length.
Donghao Zhou, Guisheng Liu, Hao Yang +9 authors
OmniShow is an end-to-end framework for human-object interaction video generation that effectively integrates multiple modalities through unified conditioning and attention mechanisms while addressing data scarcity via decoupled training strategies.
Xiaomeng Hu, Yinger Zhang, Fei Huang +7 authors
OccuBench presents a comprehensive benchmark for evaluating AI agents across 100 professional domains using Language World Models to simulate real-world environments with controlled fault injection.
Dinging Li, Yingxiu Zhao, Xinrui Cheng +16 authors
SpatialEvo is a self-evolving framework for 3D spatial reasoning that uses deterministic geometric environments to provide objective feedback, enabling efficient training without relying on model consensus.
Rui Xu, Dafei Qin, Kaichun Qiao +8 authors
SATO introduces a novel token ordering strategy for autoregressive transformers that preserves edge flow and semantic layout in mesh generation through triangle strip-based sequences.
Zile Wang, Zexiang Liu, Jaixing Li +20 authors
Matrix-Game 3.0 enhances interactive video generation through memory-augmented diffusion models that achieve real-time 720p video synthesis with long-term temporal consistency.
Zeyue Tian, Binxin Yang, Zhaoyang Liu +8 authors
Audio-Omni presents the first end-to-end framework unifying audio generation and editing across sound, music, and speech domains using a frozen multimodal language model and trainable diffusion transformer architecture.
Luozheng Qin, Jia Gong, Qian Qiao +6 authors
Uni-ViGU presents a generation-centric approach to unified multimodal video understanding and generation by extending video generation as a foundation through unified flow matching and bidirectional training mechanisms.
Dewei Zhou, You Li, Zongxin Yang +1 authors
A multimodal diffusion-based model called RefineAnything is presented for region-specific image refinement that preserves backgrounds while enhancing local details, using a focus-and-refine strategy and boundary-aware loss functions.
Tianchang Shen, Sherwin Bahmani, Kai He +12 authors
Lyra 2.0 enables large-scale 3D scene creation through persistent video generation that addresses spatial forgetting and temporal drifting issues in long-horizon video models.
Songlin Yang, Xianghao Kong, Anyi Rao
Unified multimodal models suffer from pseudo-unification due to asymmetric encoding and split response patterns, requiring consistent information flow for genuine multimodal synergy.
Lin-Zhuo Chen, Jian Gao, Yihang Chen +8 authors
LingBot-Map is a feed-forward 3D foundation model that reconstructs scenes from video streams using a geometric context transformer architecture with specialized attention mechanisms for coordinate grounding, dense geometric cues, and long-range drift correction, achieving stable real-time performance at 20 FPS.
NVIDIA, Aakshita Chandiramani, Aaron Blakeman +543 authors
Nemotron 3 Super is a 120 billion parameter hybrid Mamba-Attention Mixture-of-Experts model pre-trained in NVFP4 with LatentMoE architecture and MTP layers for accelerated inference, achieving superior throughput compared to existing models.
Han Li, Yifan Yao, Letian Zhu +13 authors
CodeTracer is a tracing architecture that analyzes code agent execution by reconstructing state transitions and localizing failures in complex multi-stage workflows.
CocoaBench Team, Shibo Hao, Zhining Zhang +29 authors
A new benchmark called CocoaBench evaluates unified digital agents on complex, multi-capability tasks requiring vision, search, and coding integration, revealing significant room for improvement in current agent systems.
Qianqian Xie, Qingheng Xiong, He Zhu +16 authors
DR$^{3}$-Eval is a benchmark for evaluating deep research agents on multimodal, multi-file report generation, featuring a realistic simulation of web environments and a comprehensive evaluation framework.
Zixian Huang, Kaichen Yang, Xu Huang +6 authors
Teacher-student cooperation data synthesis framework addresses stylistic divergence in synthetic data for improved model fine-tuning performance.
Guoxin Chen, Jie Chen, Lei Chen +7 authors
AiScientist enables autonomous long-horizon ML research engineering by combining hierarchical orchestration with durable state management, achieving superior performance on benchmark tasks through structured coordination and persistent project artifacts.
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