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。
MiniMax, Aili Chen, Aonian Li +124 authors
A hybrid-attention reasoning model called MiniMax-M1, featuring a Mixture-of-Experts architecture and lightning attention mechanism, is introduced for efficient long-input processing and reinforcement learning.
50 篇论文 · 按点赞排序
Xueqing Peng, Lingfei Qian, Yan Wang +41 authors
MultiFinBen is a multilingual and multimodal benchmark for financial domain tasks, evaluating LLMs across modalities and linguistic settings, revealing challenges in complex cross-lingual and multimodal financial reasoning.
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.
Yuhao Zhou, Yiheng Wang, Xuming He +24 authors
Scientists' First Exam (SFE) benchmark assesses scientific cognitive capacities of Multimodal Large Language Models through perception, understanding, and comparative reasoning.
Zhen Li, Chuanhao Li, Xiaofeng Mao +17 authors
Sekai, a worldwide video dataset with comprehensive annotations, is introduced to support world exploration applications, enhancing video generation models.
King Zhu, Hanhao Li, Siwei Wu +12 authors
Systematic exploration of test-time scaling methods in large language agents reveals that computational scaling improves performance, especially through parallel sampling, sequential revision, effective verification, and increased rollout diversity.
Chenlong Wang, Yuanning Feng, Dongping Chen +3 authors
NoWait suppresses explicit self-reflection tokens during inference to enhance efficiency in multimodal reasoning without reducing model utility.
Dongwei Jiang, Alvin Zhang, Andrew Wang +2 authors
LLMs show resistance to feedback, termed feedback friction, even under ideal conditions, and sampling-based strategies only partially mitigate this issue.
Yinghao Ma, Siyou Li, Juntao Yu +2 authors
CMI-Bench introduces a comprehensive instruction-following benchmark for audio-text LLMs to evaluate them on a diverse range of music information retrieval tasks.
Zhoujun Cheng, Shibo Hao, Tianyang Liu +21 authors
Guru, a diverse RL reasoning corpus, highlights domain-specific training needs and demonstrates improved performance in complex tasks for RL-enhanced LLMs.
Essential AI, Andrew Hojel, Michael Pust +21 authors
A large, 24-trillion-token Essential-Web v1.0 dataset annotated with a multi-category taxonomy outperforms or is competitive with existing datasets in various domains using simple filtering techniques.
Xumeng Wen, Zihan Liu, Shun Zheng +9 authors
RLVR advances machine reasoning by incentivizing correct and logical thought chains, addressing limitations identified by a more precise evaluation metric, $CoT$-$Pass@K$.
Xiaoran Liu, Zhigeng Liu, Zengfeng Huang +3 authors
This study investigates long-context performance of diffusion LLMs compared to auto-regressive LLMs, identifies their unique characteristics, and proposes LongLLaDA, a training-free method for extending context windows.
Shulin Tian, Ruiqi Wang, Hongming Guo +7 authors
Ego-R1, a reinforcement learning-based framework, uses a structured tool-augmented chain-of-thought process to reason over ultra-long egocentric videos, achieving better performance than existing methods by extending time coverage to a week.
Runpeng Yu, Qi Li, Xinchao Wang
Discrete Diffusion Language Models (dLLMs) and Discrete Diffusion Multimodal Language Models (dMLLMs) enable parallel generation and faster inference compared to autoregressive models through denoising-based strategies and full attention mechanisms.
Byung-Kwan Lee, Ryo Hachiuma, Yong Man Ro +2 authors
GenRecal, a novel distillation framework, improves performance of vision-language models by aligning feature representations across different architectures.
Nikolay Blagoev, Oğuzhan Ersoy, Lydia Yiyu Chen
A novel method, CheckFree, and its extended version CheckFree+, efficiently recover from node failures during LLM training by substituting failed stages with averaged neighboring stages or through out-of-order pipeline execution, improving convergence time over existing checkpointing methods.
Md Tanzib Hosain, Salman Rahman, Md Kishor Morol +1 authors
Xolver, a multi-agent reasoning framework, enhances large language models with persistent memory and diverse experience modalities, improving performance on complex reasoning tasks by avoiding generating solutions from scratch.
Feng He, Zijun Chen, Xinnian Liang +4 authors
ProtoReasoning enhances large reasoning models through prototypical representations, leading to improved cross-domain generalization in logical reasoning, planning, and other tasks.
Itay Nakash, George Kour, Koren Lazar +3 authors
CRAFT, a multi-agent system using policy-aware persuasive strategies, challenges policy-adherent LLM-based agents in customer service to assess and improve their robustness against adversarial attacks.
Subham Sekhar Sahoo, Justin Deschenaux, Aaron Gokaslan +3 authors
Duo improves uniform-state discrete diffusion models by transferring techniques from Gaussian diffusion, enhancing training speed and enabling fast few-step text generation.
Min-Seop Kwak, Junho Kim, Sangdoo Yun +4 authors
A diffusion-based framework generates aligned novel views of images and geometry using warping-and-inpainting with cross-modal attention distillation and proximity-based mesh conditioning, achieving high-fidelity synthesis and 3D completion.
Mido Assran, Adrien Bardes, David Fan +27 authors
A self-supervised approach combining internet video data and minimal robot interaction achieves strong performances in motion understanding, action anticipation, video question-answering, and robotic planning without task-specific training or reward.
Lijun Liu, Ruiyang Li, Zhaocheng Liu +5 authors
An RLVR framework using fine-tuned Qwen2.5-VL-7B achieves state-of-the-art performance in medical VIE with limited annotated samples, enhancing reasoning and balance between precision and recall.
Jinheng Xie, Zhenheng Yang, Mike Zheng Shou
Show-o2 leverages autoregressive modeling and flow matching within a 3D causal variational autoencoder to create unified visual representations for multimodal understanding and generation tasks.
Daixuan Cheng, Shaohan Huang, Xuekai Zhu +4 authors
Introducing an entropy-based term to the advantage function in reinforcement learning enhances exploratory reasoning in language models, leading to improved performance on complex reasoning tasks.
Dingcheng Zhen, Qian Qiao, Tan Yu +5 authors
TransDiff, combining an Autoregressive Transformer and diffusion models, achieves superior image generation performance and speed, while Multi-Reference Autoregression further enhances its quality and diversity.
Shaolei Zhang, Shoutao Guo, Qingkai Fang +2 authors
Stream-Omni, a large multimodal model, integrates text, vision, and speech by efficiently aligning modalities using sequence-dimension concatenation for vision and layer-dimension mapping for speech, achieving strong performance with less data.
Jingyuan Qi, Zhiyang Xu, Qifan Wang +1 authors
Autoregressive Retrieval Augmentation enhances image generation through context-aware patch-level retrievals, improving performance over existing methods.
Yuheng Yuan, Qiuhong Shen, Shizun Wang +2 authors
Test3R, a test-time learning technique for 3D reconstruction, enhances geometric accuracy by optimizing network consistency using self-supervised learning on image triplets.
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