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。
Wei Zhou, Jun Zhou, Haoyu Wang +16 authors
LLM-enhanced data preparation methods are transforming data-centric workflows from rule-based pipelines to prompt-driven, context-aware approaches, organized into data cleaning, integration, and enrichment tasks.
50 篇论文 · 按点赞排序
Meituan LongCat Team, Anchun Gui, Bei Li +159 authors
A 560-billion-parameter Mixture-of-Experts reasoning model achieves state-of-the-art performance on agentic benchmarks through a unified training framework combining domain-parallel expert training with fusion, along with enhancements for real-world robustness and complex reasoning.
Tengyue Xu, Zhuoyang Qian, Gaoge Liu +16 authors
Offline knowledge construction through structured methodological graphs enables more reliable and scalable autonomous scientific discovery by reducing reliance on real-time literature processing.
Robbyant Team, Zelin Gao, Qiuyu Wang +21 authors
LingBot-World is an open-source world simulator with high-fidelity dynamics, long-term memory capabilities, and real-time interactivity for diverse environments.
Dongrui Liu, Qihan Ren, Chen Qian +40 authors
AI agents face safety and security challenges from autonomous tool use and environmental interactions, requiring advanced guardrail frameworks for risk diagnosis and transparent monitoring.
Ji Zeng, Dayuan Fu, Tiantian Mi +14 authors
Agentic mid-training enables large language models to develop autonomous software engineering capabilities through specialized data synthesis techniques that bridge the gap between static training data and dynamic development environments.
Yanqi Dai, Yuxiang Ji, Xiao Zhang +3 authors
MathForge enhances mathematical reasoning in large models through a dual framework combining difficulty-aware policy optimization and multi-aspect question reformulation to address limitations in existing reinforcement learning methods.
Zengbin Wang, Xuecai Hu, Yong Wang +3 authors
A new benchmark and dataset are introduced to evaluate and improve spatial reasoning capabilities in text-to-image models through information-dense prompts and fine-tuning.
Hong Liu, Jiaqi Zhang, Chao Wang +13 authors
Embedding scaling offers superior sparsity scaling compared to expert scaling in large language models, enabling efficient inference through system optimizations and speculative decoding.
Yuhang Wang, Yuling Shi, Mo Yang +7 authors
SWE-Pruner is a self-adaptive context pruning framework for coding agents that uses task-aware pruning to reduce token usage while maintaining performance.
Zichen Wen, Boxue Yang, Shuang Chen +31 authors
Innovator-VL demonstrates that principled training design and transparent methodology can achieve strong scientific intelligence with reduced data requirements while maintaining general vision performance.
Haozhe Xie, Beichen Wen, Jiarui Zheng +4 authors
DynamicVLA addresses dynamic object manipulation challenges through a compact vision-language-action model with temporal reasoning and closed-loop adaptation, supported by a new benchmark for dynamic manipulation tasks.
Haoran Wei, Yaofeng Sun, Yukun Li
DeepSeek-OCR 2 introduces DeepEncoder V2 that dynamically reorders visual tokens based on semantic content, enabling more human-like causal reasoning in 2D image understanding through cascaded 1D causal structures.
Honglin Lin, Zheng Liu, Yun Zhu +6 authors
A large-scale multimodal reasoning dataset called MMFineReason is introduced to improve vision language models' performance through high-quality reasoning annotations and demonstrates superior parameter efficiency in fine-tuned models.
Bin Yu, Shijie Lian, Xiaopeng Lin +8 authors
TwinBrainVLA addresses the tension between semantic understanding and motor skills in robot control by coordinating a generalist vision-language model with a specialist model through an asymmetric mixture-of-transformers mechanism.
Chenyu Mu, Xin He, Qu Yang +13 authors
A novel end-to-end agentic framework translates dialogue into cinematic videos through specialized agents that generate and orchestrate video content while maintaining narrative coherence.
Yufeng Zhong, Lei Chen, Xuanle Zhao +7 authors
OCRVerse is a novel end-to-end OCR method that unifies text-centric and vision-centric approaches through comprehensive data engineering and a two-stage SFT-RL training framework with domain-specific reward strategies.
Jonas Hübotter, Frederike Lübeck, Lejs Behric +8 authors
Self-Distillation Policy Optimization (SDPO) enhances reinforcement learning with verifiable rewards by utilizing rich textual feedback to improve sample efficiency and accuracy in language model training.
Wei Wu, Fan Lu, Yunnan Wang +22 authors
A Vision-Language-Action model trained on extensive real-world robotic data demonstrates superior performance and generalization across multiple platforms while offering enhanced efficiency through optimized training infrastructure.
Mingyang Song, Haoyu Sun, Jiawei Gu +4 authors
AdaReasoner enables multimodal models to learn tool usage as a general reasoning skill through scalable data curation, reinforcement learning for tool selection, and adaptive learning mechanisms that improve performance on complex visual reasoning tasks.
Xian Shi, Xiong Wang, Zhifang Guo +10 authors
The Qwen3-ASR family introduces speech recognition models with language identification capabilities and a non-autoregressive forced alignment model, achieving state-of-the-art performance and efficient processing.
Zhixiang Wei, Yi Li, Zhehan Kan +38 authors
Youtu-VL addresses limitations in Vision-Language Models by introducing a unified autoregressive supervision paradigm that treats visual signals as target outputs rather than passive inputs, enabling improved multimodal comprehension and vision-centric task performance.
Shobhita Sundaram, John Quan, Ariel Kwiatkowski +3 authors
A self-improvement framework enables pretrained language models to generate automated curricula for solving previously unsolvable problems by leveraging latent knowledge and meta-reinforcement learning.
Zihao Huang, Jundong Zhou, Xingwei Qu +2 authors
ConceptMoE dynamically allocates computation by merging similar tokens into concept representations, improving both performance and efficiency in large language models through adaptive processing and reduced attention computation.
Honglin Lin, Chonghan Qin, Zheng Liu +7 authors
Scientific image synthesis using logic-driven frameworks like ImgCoder improves multimodal reasoning by addressing visual-logic divergence through structured generation and evaluation benchmarks.
Idan Shenfeld, Mehul Damani, Jonas Hübotter +1 authors
Self-Distillation Fine-Tuning enables on-policy learning from demonstrations, reducing catastrophic forgetting and allowing continuous skill accumulation in foundation models.
Zirui Wang, Junyi Zhang, Jiaxin Ge +9 authors
Modern vision-language models exhibit significant challenges in multi-step visual interaction tasks, particularly in long-horizon perception-memory-action integration, with performance declining when handling unbounded historical contexts.
Yinger Zhang, Shutong Jiang, Renhao Li +6 authors
DeepPlanning benchmark addresses limitations of current LLM planning assessments by introducing complex, real-world tasks requiring both global optimization and local constraint reasoning.
Bin Lin, Zongjian Li, Yuwei Niu +9 authors
Finite Scalar Quantization with improved activation mapping enables unified modeling of discrete and continuous image generation approaches, revealing optimal representation balance and performance characteristics.
Zecheng Tang, Quantong Qiu, Yi Yang +6 authors
Elastic Attention enables dynamic adjustment of attention sparsity during inference by integrating a lightweight Attention Router into pretrained models, achieving efficient long-context processing.
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