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。
606 篇论文 · 按点赞排序
Sangmin Bae, Yujin Kim, Reza Bayat +8 authors
Mixture-of-Recursions (MoR) combines parameter sharing and adaptive computation in a Recursive Transformer to improve efficiency, reduce memory usage, and enhance performance across different model scales.
Zafir Stojanovski, Oliver Stanley, Joe Sharratt +4 authors
Reasoning Gym provides a library of reasoning environments with verifiable rewards and procedural data generation for reinforcement learning, enabling the evaluation and training of reasoning models at varying difficulty levels.
Mingxuan Du, Benfeng Xu, Chiwei Zhu +2 authors
DeepResearch Bench offers a benchmark framework to evaluate the capabilities of Deep Research Agents in terms of research quality and information retrieval accuracy across multiple fields.
Chenze Shao, Darren Li, Fandong Meng +1 authors
Continuous Autoregressive Language Models (CALM) improve language model efficiency by predicting continuous vectors instead of discrete tokens, reducing computational cost while maintaining performance.
Mohammad Zbib, Hasan Abed Al Kader Hammoud, Sina Mukalled +5 authors
AraLingBench evaluates Arabic and bilingual LLMs' linguistic competence using a benchmark with expert-designed questions across grammar, morphology, spelling, reading comprehension, and syntax, revealing gaps between surface proficiency and true comprehension.
Ruibin Yuan, Hanfeng Lin, Shuyue Guo +54 authors
YuE, a family of open foundation models based on LLaMA2, can generate long-form music with aligned lyrics, coherent structure, and appropriate accompaniment using innovative techniques in next-token prediction, conditioning, and pre-training.
Eduard Allakhverdov, Elizaveta Goncharova, Andrey Kuznetsov
A method using autoencoder and Gumbel-Softmax selection identifies and retains only the most informative visual tokens, enabling efficient multimodal pruning with minimal performance loss.
Yuchao Gu, Weijia Mao, Mike Zheng Shou
FAR, an autoregressive video model, achieves state-of-the-art results by balancing locality and long-range context with FlexRoPE and long short-term context modeling, addressing challenges in long-context video generation.
Runnan Fang, Shihao Cai, Baixuan Li +13 authors
A scalable framework and two-phase fine-tuning strategy enhance function-calling capabilities of agents in diverse environments, improving performance on agentic benchmarks.
Zhiyuan Zeng, Jiashuo Liu, Siyuan Chen +27 authors
FutureX is a dynamic, live benchmark for evaluating LLM agents in future prediction tasks, addressing challenges in real-time updates and data contamination.
Yue Liao, Pengfei Zhou, Siyuan Huang +11 authors
Genie Envisioner integrates policy learning, evaluation, and simulation using a video diffusion model and neural simulator for instruction-driven robotic manipulation.
Zhengliang Shi, Yiqun Chen, Haitao Li +23 authors
Deep Research systems integrate LLMs with external tools to enhance problem-solving capabilities, involving query planning, information acquisition, memory management, and answer generation.
Xiaoran Liu, Ruixiao Li, Mianqiu Huang +10 authors
The survey examines the advancements and challenges in long-context Large Language Models (LLMs), exploring the architecture, infrastructure, training, and evaluation technologies needed to extend their context length and address the inherent trade-offs.
Jie Wu, Yu Gao, Zilyu Ye +9 authors
RewardDance is a scalable reward modeling framework that aligns with VLM architectures, enabling effective scaling of RMs and resolving reward hacking issues in generation models.
Ruizhong Qiu, Gaotang Li, Tianxin Wei +2 authors
SAFFRON, a novel inference scaling paradigm, enhances LLM safety by reducing reward model evaluations through a multifurcation reward model and other optimizations.
Junyi Hou, Andre Lin Huikai, Nuo Chen +2 authors
PaperDebugger is an in-editor academic writing assistant that integrates large language models, enabling direct interaction within LaTeX editors for document state management, revision, and literature search.
Qiushi Sun, Mukai Li, Zhoumianze Liu +11 authors
A hybrid safety detection framework combining formal verification and VLM-based contextual assessment improves the detection of unsafe operations in mobile agents.
Nanye Ma, Shangyuan Tong, Haolin Jia +8 authors
Inference-time scaling of diffusion models improves sample quality beyond increasing denoising steps through better noise selection and verification.
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.
Wenqiang Sun, Haiyu Zhang, Haoyuan Wang +7 authors
WorldPlay is a streaming video diffusion model that achieves real-time, interactive world modeling with long-term geometric consistency by using a Dual Action Representation, Reconstituted Context Memory, and Context Forcing.
Jingwei Zuo, Maksim Velikanov, Ilyas Chahed +24 authors
Falcon-H1, a new series of large language models with a hybrid architecture combining Transformer-based attention and State Space Models, achieves state-of-the-art performance and efficiency across various tasks and sizes.
Yumin Choi, Jinheon Baek, Sung Ju Hwang
A meta-learning framework for optimizing system prompts in Large Language Models (LLMs) improves generalization across diverse tasks and datasets.
An Yang, Bowen Yu, Chengyuan Li +25 authors
The Qwen2.5-1M series models extend context length to 1 million tokens with enhanced long-context capabilities, employing techniques like long data synthesis and progressive pre-training, and are supported by an open-source inference framework with sparse attention and kernel optimizations.
Zhengyang Tang, Ziniu Li, Zhenyang Xiao +8 authors
SCRIT, a self-evolving critique framework, enhances LLMs' critique capabilities using synthetic data and self-validation, achieving significant improvements in critique-correction and error identification benchmarks.
Sambal Shikhar, Mohammed Irfan Kurpath, Sahal Shaji Mullappilly +5 authors
LLMVoX is a lightweight, LLM-agnostic TTS system that decouples speech synthesis from model processing, ensuring high-quality speech with low latency and supporting seamless multimodal dialogue.
Ye Tian, Ling Yang, Jiongfan Yang +10 authors
A parallel multimodal diffusion framework, MMaDA-Parallel, enhances cross-modal alignment and semantic consistency in thinking-aware image synthesis by addressing error propagation issues in sequential approaches.
Chunshi Wang, Junliang Ye, Yunhan Yang +6 authors
We introduce Part-X-MLLM, a native 3D multimodal large language model that unifies diverse 3D tasks by formulating them as programs in a structured, executable grammar. Given an RGB point cloud and a natural language prompt, our model autoregressively generates a single, coherent token sequence encoding part-level bounding boxes, semantic descriptions, and edit commands. This structured output serves as a versatile interface to drive downstream geometry-aware modules for part-based generation and editing. By decoupling the symbolic planning from the geometric synthesis, our approach allows any compatible geometry engine to be controlled through a single, language-native frontend. We pre-train a dual-encoder architecture to disentangle structure from semantics and instruction-tune the model on a large-scale, part-centric dataset. Experiments demonstrate that our model excels at producing high-quality, structured plans, enabling state-of-the-art performance in grounded Q\&A, compositional generation, and localized editing through one unified interface. Project page: https://chunshi.wang/Part-X-MLLM/
Mikhail Salnikov, Dmitrii Korzh, Ivan Lazichny +7 authors
LLMs exhibit significant geopolitical biases in their interpretation of historical events, and simple debiasing methods have limited effectiveness; a novel dataset for further research is provided.
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.
Mo Yu, Tsz Ting Chung, Chulun Zhou +8 authors
A benchmark called PRELUDE evaluates long-context understanding by assessing the consistency of prequel stories with original books, revealing significant challenges for models compared to humans.
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