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。
Qingxiu Dong, Li Dong, Yao Tang +4 authors
Reinforcement Pre-Training (RPT) improves language model accuracy through reinforcement learning and offers a scalable method for leveraging text data for general-purpose RL.
50 篇论文 · 按点赞排序
Sergey Pletenev, Maria Marina, Nikolay Ivanov +6 authors
EverGreenQA, a multilingual QA dataset with evergreen labels, is introduced to benchmark LLMs on temporality encoding and assess their performance through verbalized judgments and uncertainty signals.
Pengyi Li, Matvey Skripkin, Alexander Zubrey +2 authors
Reinforcement Learning via Self-Confidence (RLSC) improves large language model accuracy using the model's confidence as a reward signal, eliminating the need for human labels or reward engineering.
LASA Team, Weiwen Xu, Hou Pong Chan +16 authors
A medical-specialized multimodal large language model, Lingshu, is introduced with enhanced data curation and reinforcement learning to address limitations in medical applications.
Yu Gao, Haoyuan Guo, Tuyen Hoang +41 authors
Seedance 1.0 offers high-performance video generation by integrating advanced data curation, efficient architecture, post-training optimization, and model acceleration, resulting in superior quality and speed.
MiniCPM Team, Chaojun Xiao, Yuxuan Li +72 authors
MiniCPM4, a highly efficient large language model for end-side devices, achieves superior performance using innovations in sparse attention, pre-training datasets, training algorithms, and inference systems.
Yu Sun, Xingyu Qian, Weiwen Xu +7 authors
ReasonMed, a large medical reasoning dataset, enhances the accuracy of medical question answering models by combining detailed reasoning paths with concise summaries, setting new benchmarks for model performance.
Yuchen Lin, Chenguo Lin, Panwang Pan +4 authors
PartCrafter is a unified 3D generative model that synthesizes multiple semantically meaningful 3D meshes from a single image using a compositional latent space and hierarchical attention mechanism.
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.
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.
Mistral-AI, Abhinav Rastogi, Albert Q. Jiang +97 authors
Magistral, a scalable reinforcement learning pipeline, demonstrates that RL can enhance multimodal understanding and instruction following in large language models without requiring existing RL traces.
Xinyu Yang, Yuwei An, Hongyi Liu +2 authors
Multiverse, a generative model enabling native parallel text generation, achieves performance comparable to autoregressive models through a three-stage MapReduce paradigm and efficient attention mechanisms.
Lianghong Guo, Yanlin Wang, Caihua Li +6 authors
A pipeline named SWE-Factory automates the creation and validation of GitHub issue resolution datasets for training and evaluating Large Language Models, using SWE-Builder for environment setup, exit-code-based grading, and automated fail2pass validation.
Zhenran Xu, Yiyu Wang, Xue Yang +5 authors
ComfyUI-R1, a large reasoning model for automated workflow generation, demonstrates superior performance in creating AI art workflows through long chain-of-thought reasoning and reinforcement learning.
Yongsen Mao, Junhao Zhong, Chuan Fang +5 authors
SpatialLM, a multimodal large language model, processes 3D point cloud data to generate structured scene understanding outputs, achieving state-of-the-art performance in layout estimation and competitive results in 3D object detection.
Shanchuan Lin, Ceyuan Yang, Hao He +6 authors
Autoregressive adversarial post-training transforms a pre-trained latent video diffusion model into a real-time, interactive video generator with efficient one-step generation and reduced error accumulation.
Jaewon Min, Jin Hyeon Kim, Paul Hyunbin Cho +6 authors
The proposed Text-Aware Image Restoration (TAIR) system integrates a multi-task diffusion framework with a text-spotting module to enhance both image recovery and textual fidelity, outperforming existing diffusion-based methods.
Jingjing Chang, Yixiao Fang, Peng Xing +6 authors
OneIG-Bench is a comprehensive benchmark framework for evaluating text-to-image models across multiple dimensions including reasoning, text rendering, and diversity.
Shenshen Li, Kaiyuan Deng, Lei Wang +6 authors
A new data selection paradigm, Reasoning Activation Potential (RAP), enhances multi-modal reasoning in large language models using minimal high-value datasets, improving performance and reducing computational costs.
Haoyuan Shi, Yunxin Li, Xinyu Chen +3 authors
AniMaker, a multi-agent framework using MCTS-Gen and AniEval, generates coherent storytelling videos from text input, outperforming existing models with better quality and efficiency.
Ananth Muppidi, Abhilash Nandy, Sambaran Bandyopadhyay
A new method using input-dependent soft prompting with a self-attention mechanism improves parameter-efficient fine-tuning for large language models, enhancing zero-shot domain transfer.
Yuanpeng Tu, Hao Luo, Xi Chen +3 authors
PlayerOne is an egocentric realistic world simulator that constructs and generates videos from user-captured images, using a coarse-to-fine training pipeline and advanced motion injection and reconstruction frameworks.
Zikui Cai, Andrew Wang, Anirudh Satheesh +10 authors
MORSE-500, a video benchmark with 500 scripted clips, evaluates multimodal reasoning across six categories, highlighting performance gaps in abstract and planning tasks.
Dianyi Wang, Wei Song, Yikun Wang +4 authors
Autoregressive Semantic Visual Reconstruction (ASVR) improves multimodal understanding by focusing on semantic reconstruction rather than raw visual appearance, enhancing performance across various benchmarks.
Shunian Chen, Xinyuan Xie, Zheshu Chen +5 authors
A novel two-stage pipeline using specialized pretrained models and a large language model enhances audio caption quality by integrating diverse multimodal cues and contextual information.
Pooneh Mousavi, Gallil Maimon, Adel Moumen +18 authors
A systematic review and benchmark of discrete audio tokenizers across speech, music, and general audio domains is presented, covering their taxonomy, evaluation metrics, and limitations.
Inclusion AI, Biao Gong, Cheng Zou +55 authors
Ming-Omni is a unified multimodal model with dedicated encoders and modality-specific routers that can process images, text, audio, and video, and performs tasks like speech and image generation, context-aware chatting, and versatile image editing.
Xun Huang, Zhengqi Li, Guande He +2 authors
Self Forcing, a novel training method for autoregressive video diffusion models, reduces exposure bias and improves generation quality through holistic video-level supervision and efficient caching mechanisms.
Jiashuo Yu, Yue Wu, Meng Chu +14 authors
VRBench is a long narrative video benchmark designed to evaluate models' multi-step reasoning and procedural validity through human-labeled question-answering pairs and a human-AI collaborative framework with a multi-phase evaluation pipeline.
Yang Liu, Jiaqi Li, Zilong Zheng
RuleReasoner enhances rule-based reasoning in small models through dynamic domain sampling, achieving superior performance and efficiency compared to large models.
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