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。
Xinyu Guan, Li Lyna Zhang, Yifei Liu +5 authors
rStar-Math enhances small language models' math reasoning capabilities through Monte Carlo Tree Search and self-evolution, achieving state-of-the-art performance on various benchmarks without distillation from larger models.
50 篇论文 · 按点赞排序
Xiaoxi Li, Guanting Dong, Jiajie Jin +5 authors
Search-o1 enhances large reasoning models with an agentic retrieval-augmented generation mechanism and a Reason-in-Documents module to improve performance on complex reasoning tasks.
Jian Hu
REINFORCE++ is an enhanced reinforcement learning algorithm that combines PPO techniques without a critic network, offering simplicity, stability, and efficiency.
Yiwen Huang, Aaron Gokaslan, Volodymyr Kuleshov +1 authors
A new regularization approach and simplified architecture for GANs, named R3GAN, surpasses existing methods including StyleGAN2 and diffusion models across multiple datasets.
Violet Xiang, Charlie Snell, Kanishk Gandhi +11 authors
The Meta Chain-of-Thought framework extends traditional CoT by explicitly modeling reasoning, enabling more powerful and human-like reasoning in LLMs through process supervision, synthetic data generation, and reinforcement learning.
Samuel Schmidgall, Yusheng Su, Ze Wang +6 authors
Agent Laboratory, an LLM-based framework, accelerates scientific research by automating the entire process from idea to report, resulting in high-quality outputs with reduced expenses.
NVIDIA, Niket Agarwal, Arslan Ali +75 authors
The Cosmos World Foundation Model Platform provides a framework for developing customized world models for Physical AI by leveraging video curation, pre-trained models, and tokenizers.
Ethem Yağız Çalık, Talha Rüzgar Akkuş
Advancements in enhancing natural language understanding, conversational coherence, and emotional intelligence in large language models improve user interactions and expand AI applications, while future research will address ethical implications and biases.
Rui Xie, Yinhong Liu, Penghao Zhou +7 authors
A novel approach using text-to-video models with local information enhancement and dynamic frequency loss improves spatio-temporal quality in real-world video super-resolution.
Siyuan Huang, Liliang Chen, Pengfei Zhou +7 authors
EnerVerse enhances robotic manipulation by generating future space using convolutional and bidirectional attention, sparse memory, and generated multi-camera data.
Shaolei Zhang, Qingkai Fang, Zhe Yang +1 authors
LLaVA-Mini, a unified large multimodal model, significantly reduces computational cost and latency by compressing vision tokens to a single token before feeding them to the LLM backbone, outperforming previous models.
Haobo Yuan, Xiangtai Li, Tao Zhang +7 authors
Sa2VA unifies image and video understanding by integrating SAM-2 and LLaVA into a shared token space, achieving state-of-the-art performance across multiple tasks.
Ruilin Luo, Zhuofan Zheng, Yifan Wang +5 authors
A three-module synthesis strategy and data generation approach improve Chain-of-Thought reasoning in multimodal mathematical tasks, boosting model performance and out-of-distribution verification capabilities.
Chaoyou Fu, Haojia Lin, Xiong Wang +12 authors
A multi-stage training method enhances multimodal large language models with both visual and speech capabilities, enabling real-time interaction without separate ASR and TTS modules.
Yixin Ji, Juntao Li, Hai Ye +4 authors
Test-time computing scaling enhances complex reasoning in models, improving robustness, generalization, and problem-solving from System-1 to System-2 thinking through various mechanisms.
Wenyi Hong, Yean Cheng, Zhuoyi Yang +6 authors
MotionBench evaluates and improves vision language models' fine-grained motion comprehension through redesigned benchmarking and an efficient Through-Encoder Fusion method.
Beichen Zhang, Yuhong Liu, Xiaoyi Dong +6 authors
BoostStep improves large language models' reasoning quality in math problems by aligning granularity and providing relevant in-context learning examples, enhancing performance and integrating with Monte Carlo Tree Search methods.
Jathushan Rajasegaran, Ilija Radosavovic, Rahul Ravishankar +3 authors
Autoregressive pre-training on a massive dataset of videos and images using transformer models yields competitive performance across various tasks and exhibits scaling behavior similar to language models.
Ziming Luo, Zonglin Yang, Zexin Xu +2 authors
LLMs are transforming scientific research across hypothesis discovery, experiment planning, writing, and reviewing, offering new methodologies and benchmarks while highlighting challenges for future research.
Rui Qian, Shuangrui Ding, Xiaoyi Dong +5 authors
Dispider, a system for active real-time interaction with video LLMs, disentangles and asynchronously processes video for timely, contextually accurate, and efficient responses in streaming scenarios.
Yifan Du, Zikang Liu, Yifan Li +7 authors
Fine-tuning multimodal large language models with long-form textual reasoning data enhances their slow-thinking capabilities, demonstrating the transferability of these capacities across modalities.
Steven Au, Cameron J. Dimacali, Ojasmitha Pedirappagari +7 authors
PGraphRAG uses user-centric knowledge graphs to enhance personalization in text generation, outperforming existing methods especially in scenarios with sparse data.
Xingwu Sun, Shuaipeng Li, Ruobing Xie +13 authors
Floating-point quantization training of LLMs is explored to optimize performance and reduce costs, revealing critical insights into bit allocation, data size impacts, and precision-performance relationships.
Shaoyuan Xie, Lingdong Kong, Yuhao Dong +5 authors
DriveBench evaluates VLMs for autonomous driving, revealing limitations in visual grounding and multi-modal reasoning, and proposing improved metrics for reliability and interpretability.
Luozhou Wang, Yijun Li, Zhifei Chen +5 authors
TransPixar extends pretrained video models to generate RGBA videos using a diffusion transformer with LoRA-based fine-tuning, ensuring consistency between RGB and alpha channels.
Yuhang Liu, Pengxiang Li, Zishu Wei +7 authors
InfiGUIAgent, an MLLM-based GUI Agent, enhances multi-step reasoning and GUI interaction through a two-stage supervised fine-tuning pipeline.
Hao Zheng, Xinyan Guan, Hao Kong +6 authors
PPTAgent, a two-stage approach, improves presentation generation by analyzing reference presentations and ensuring structural and content consistency, outperforming traditional methods across content, design, and coherence.
Zekai Gu, Rui Yan, Jiahao Lu +9 authors
Diffusion as Shader (DaS) achieves versatile video control by leveraging 3D tracking signals, enabling tasks like mesh-to-video generation, camera control, motion transfer, and object manipulation.
Guy Yariv, Yuval Kirstain, Amit Zohar +5 authors
The proposed two-stage framework for Image-to-Video generation uses mask-based motion trajectories and object-level attention to produce realistic and semantically consistent videos from text descriptions.
Ollie Liu, Sami Jaghouar, Johannes Hagemann +4 authors
A large autoregressive transformer model is pretrained on diverse metagenomic sequences for pandemic monitoring and pathogen detection, achieving high performance on genomic benchmarks.
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