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。
Zhenda Xie, Yixuan Wei, Huanqi Cao +16 authors
Manifold-Constrained Hyper-Connections (mHC) stabilize and scale residual connection architectures by restoring identity mapping properties through manifold projection and infrastructure optimization.
50 篇论文 · 按点赞排序
Shih-Yang Liu, Xin Dong, Ximing Lu +10 authors
Multi-reward reinforcement learning suffers from reward normalization collapse in GRPO, which GDPO addresses by decoupling reward normalization for improved training stability and performance across reasoning tasks.
Chengwen Liu, Xiaomin Yu, Zhuoyue Chang +13 authors
VideoDR benchmark enables video question answering by combining cross-frame visual extraction, web retrieval, and multi-hop reasoning in open-domain settings.
Tianxin Wei, Ting-Wei Li, Zhining Liu +26 authors
Agentic reasoning redefines large language models as autonomous agents capable of planning, acting, and learning through continuous interaction in dynamic environments across single-agent and multi-agent frameworks.
Liang Chen, Weichu Xie, Yiyan Liang +26 authors
Current multimodal large language models exhibit significant gaps in fundamental visual understanding compared to human children, as demonstrated by the BabyVision benchmark.
Ailin Huang, Chengyuan Yao, Chunrui Han +90 authors
STEP3-VL-10B achieves superior multimodal performance through unified pre-training with a language-aligned Perception Encoder and Qwen3-8B decoder, combined with scaled post-training and Parallel Coordinated Reasoning for efficient large-scale visual reasoning.
Yoav HaCohen, Benny Brazowski, Nisan Chiprut +26 authors
LTX-2 is an open-source audiovisual diffusion model that generates synchronized video and audio content using a dual-stream transformer architecture with cross-modal attention and classifier-free guidance.
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.
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.
Yuxiang Ji, Yong Wang, Ziyu Ma +6 authors
Large vision-language models are enhanced for image geolocalization by incorporating map-based reasoning and agent-in-the-map loop optimization, achieving superior accuracy compared to existing models.
Fengkai Yang, Zherui Chen, Xiaohan Wang +10 authors
Group-based reinforcement learning from verifier rewards suffers from biased advantage estimation that underestimates hard prompts and overestimates easy prompts, which is addressed through a history-aware adaptive difficulty weighting method that improves performance on mathematical reasoning benchmarks.
Junru Lu, Jiarui Qin, Lingfeng Qiao +35 authors
Youtu-LLM is a lightweight language model optimized for computational efficiency and agentic intelligence through a compact architecture, STEM-focused training curriculum, and scalable mid-training strategies for planning and reasoning tasks.
Yu Wang, Yi Wang, Rui Dai +4 authors
Urban socio-semantic segmentation is achieved through a vision-language model framework that combines cross-modal recognition and multi-stage reasoning with reinforcement learning optimization.
Zhiyuan Hu, Yucheng Wang, Yufei He +7 authors
Reinforcement learning for large language models is enhanced by a rollout-level objective that rewards rare high-level reasoning strategies, improving diverse solution discovery without sacrificing initial performance.
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.
Yuxue Yang, Lue Fan, Ziqi Shi +3 authors
NeoVerse is a scalable 4D world model that enables pose-free reconstruction and novel-trajectory video generation from monocular videos with state-of-the-art performance.
Yibo Wang, Lei Wang, Yue Deng +7 authors
DeepResearchEval presents an automated framework for creating complex research tasks and evaluating them through agent-based methods that adapt to task specifics and verify facts without relying on citations.
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.
Yuchen Shi, Yuzheng Cai, Siqi Cai +15 authors
Existing Large Language Model (LLM) agent frameworks face two significant challenges: high configuration costs and static capabilities. Building a high-quality agent often requires extensive manual effort in tool integration and prompt engineering, while deployed agents struggle to adapt to dynamic environments without expensive fine-tuning. To address these issues, we propose Youtu-Agent, a modular framework designed for the automated generation and continuous evolution of LLM agents. Youtu-Agent features a structured configuration system that decouples execution environments, toolkits, and context management, enabling flexible reuse and automated synthesis. We introduce two generation paradigms: a Workflow mode for standard tasks and a Meta-Agent mode for complex, non-standard requirements, capable of automatically generating tool code, prompts, and configurations. Furthermore, Youtu-Agent establishes a hybrid policy optimization system: (1) an Agent Practice module that enables agents to accumulate experience and improve performance through in-context optimization without parameter updates; and (2) an Agent RL module that integrates with distributed training frameworks to enable scalable and stable reinforcement learning of any Youtu-Agents in an end-to-end, large-scale manner. Experiments demonstrate that Youtu-Agent achieves state-of-the-art performance on WebWalkerQA (71.47\%) and GAIA (72.8\%) using open-weight models. Our automated generation pipeline achieves over 81\% tool synthesis success rate, while the Practice module improves performance on AIME 2024/2025 by +2.7\% and +5.4\% respectively. Moreover, our Agent RL training achieves 40\% speedup with steady performance improvement on 7B LLMs, enhancing coding/reasoning and searching capabilities respectively up to 35\% and 21\% on Maths and general/multi-hop QA benchmarks.
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.
Muxi Diao, Lele Yang, Wuxuan Gong +6 authors
Entropy-Adaptive Fine-Tuning addresses catastrophic forgetting in supervised fine-tuning by using token-level entropy to distinguish uncertainty from knowledge conflict, enabling better preservation of general capabilities.
Tu Hu, Ronghao Chen, Shuo Zhang +9 authors
Controlled Self-Evolution method improves code generation through diversified initialization, feedback-guided genetic evolution, and hierarchical memory to enhance exploration efficiency and solution quality.
Chulun Zhou, Chunkang Zhang, Guoxin Yu +4 authors
Multi-step retrieval-augmented generation (RAG) has become a widely adopted strategy for enhancing large language models (LLMs) on tasks that demand global comprehension and intensive reasoning. Many RAG systems incorporate a working memory module to consolidate retrieved information. However, existing memory designs function primarily as passive storage that accumulates isolated facts for the purpose of condensing the lengthy inputs and generating new sub-queries through deduction. This static nature overlooks the crucial high-order correlations among primitive facts, the compositions of which can often provide stronger guidance for subsequent steps. Therefore, their representational strength and impact on multi-step reasoning and knowledge evolution are limited, resulting in fragmented reasoning and weak global sense-making capacity in extended contexts. We introduce HGMem, a hypergraph-based memory mechanism that extends the concept of memory beyond simple storage into a dynamic, expressive structure for complex reasoning and global understanding. In our approach, memory is represented as a hypergraph whose hyperedges correspond to distinct memory units, enabling the progressive formation of higher-order interactions within memory. This mechanism connects facts and thoughts around the focal problem, evolving into an integrated and situated knowledge structure that provides strong propositions for deeper reasoning in subsequent steps. We evaluate HGMem on several challenging datasets designed for global sense-making. Extensive experiments and in-depth analyses show that our method consistently improves multi-step RAG and substantially outperforms strong baseline systems across diverse tasks.
Weixun Wang, XiaoXiao Xu, Wanhe An +85 authors
The Agentic Learning Ecosystem (ALE) introduces a principled infrastructure for agent development, combining post-training optimization, sandbox environments, and policy alignment to enhance long-horizon training stability and performance in real-world tasks.
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.
Jing Xiong, Qi Han, Yunta Hsieh +11 authors
MMFormalizer enables multimodal autoformalization by integrating visual perception with formal mathematical reasoning, supporting complex physical domains from classical mechanics to quantum mechanics.
Hao Yu, Haotong Lin, Jiawei Wang +7 authors
InfiniDepth represents depth as neural implicit fields using a local implicit decoder, enabling continuous 2D coordinate querying for arbitrary-resolution depth estimation and superior performance in fine-detail regions.
Alex L. Zhang, Tim Kraska, Omar Khattab
We study allowing large language models (LLMs) to process arbitrarily long prompts through the lens of inference-time scaling. We propose Recursive Language Models (RLMs), a general inference strategy that treats long prompts as part of an external environment and allows the LLM to programmatically examine, decompose, and recursively call itself over snippets of the prompt. We find that RLMs successfully handle inputs up to two orders of magnitude beyond model context windows and, even for shorter prompts, dramatically outperform the quality of base LLMs and common long-context scaffolds across four diverse long-context tasks, while having comparable (or cheaper) cost per query.
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