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
Yiran Zhao, Chaoqun Liu, Yue Deng +8 authors
Babel is an open multilingual LLM that expands its parameter count through layer extension, covering numerous languages and achieving superior performance in multilingual tasks compared to other open LLMs.
Xilin Wei, Xiaoran Liu, Yuhang Zang +9 authors
VideoRoPE, a 3D position embedding that preserves spatio-temporal relationships, improves performance on video tasks by addressing limitations in previous RoPE variants.
Yujia Qin, Yining Ye, Junjie Fang +32 authors
UI-TARS, a native GUI agent model using screenshots as input, outperforms commercial models in various benchmarks through enhanced perception, unified action modeling, system-2 reasoning, and iterative training with reflective online traces.
Anlin Zheng, Xin Wen, Xuanyang Zhang +5 authors
A novel image tokenizer built on pre-trained vision foundation models improves image reconstruction, generation quality, and token efficiency, enhancing autoregressive generation and class-conditional synthesis.
Jiazhan Feng, Shijue Huang, Xingwei Qu +6 authors
ReTool, a tool-integrated learning framework, enhances reasoning models with real-time code execution and reinforcement learning, significantly improving performance in structured problem-solving tasks like mathematical reasoning.
Yan Ma, Linge Du, Xuyang Shen +7 authors
A unified reinforcement learning system, V-Triune, combines visual reasoning and perception tasks in vision-language models through a single training pipeline, achieving significant improvements across various tasks.
Dacheng Li, Shiyi Cao, Chengkun Cao +6 authors
A hybrid test-time scaling framework improves code generation coverage and accuracy across various models and domains.
Zhenting Wang, Qi Chang, Hemani Patel +8 authors
MCP-Bench evaluates large language models on complex, multi-step tasks requiring tool use, coordination, and planning across various domains.
Rulin Shao, Akari Asai, Shannon Zejiang Shen +18 authors
Reinforcement Learning with Evolving Rubrics (RLER) enables training of deep research models for long-form tasks, outperforming existing models and proprietary systems while being more cost-effective.
Nikhil Kandpal, Brian Lester, Colin Raffel +24 authors
The Common Pile v0.1 dataset, an 8-terabyte collection of openly licensed text, is used to train competitive 7 billion parameter LLMs.
Tingchen Fu, Jiawei Gu, Yafu Li +2 authors
An empirical analysis of MathIF identifies a tension between enhancing reasoning capacity and maintaining instruction adherence in large language models.
Yiren Song, Cheng Liu, Mike Zheng Shou
OmniConsistency, using large-scale Diffusion Transformers, enhances stylization consistency and generalization in image-to-image pipelines without style degradation.
Huijie Liu, Shuhao Cui, Haoxiang Cao +3 authors
A novel method, CoTyle, generates images in consistent visual styles using unique numerical style codes, filling an academic gap in code-to-style image generation.
Guo-Hua Wang, Shanshan Zhao, Xinjie Zhang +9 authors
Ovis-U1, a 3-billion-parameter unified model, integrates multimodal understanding, text-to-image generation, and image editing using a diffusion-based visual decoder and bidirectional token refiner, achieving state-of-the-art performance across various benchmarks.
Zhaoxi Chen, Tianqi Liu, Long Zhuo +6 authors
4DNeX generates high-quality dynamic 3D scene representations from a single image using a fine-tuned pretrained video diffusion model, outperforming existing methods in efficiency and generalizability.
Gleb Mezentsev, Ivan Oseledets
LLMs can generate long text segments in a single forward pass using learned embeddings, revealing a capability for multi-token generation without iterative decoding.
Yue Liu, Shengfang Zhai, Mingzhe Du +9 authors
GuardReasoner-VL enhances VLM safety through a reasoning-based guard model trained with SFT and online RL, using diverse datasets and a length-aware safety reward.
Jianyi Wang, Shanchuan Lin, Zhijie Lin +10 authors
SeedVR2 is a one-step diffusion-based video restoration model that uses adaptive window attention and feature matching loss to achieve high visual quality with reduced computational cost compared to existing methods.
Jiazhen Pan, Che Liu, Junde Wu +6 authors
MedVLM-R1 enhances medical image analysis through explicit natural language reasoning generation, achieving superior accuracy with limited data and parameters via reinforcement learning rather than supervised fine-tuning.
Yixin Ye, Zhen Huang, Yang Xiao +3 authors
LIMO, a new model, achieves high mathematical reasoning performance using minimal training data, challenging the notion that extensive datasets are necessary for complex reasoning.
Ganqu Cui, Lifan Yuan, Zefan Wang +20 authors
PRIME leverages implicit process rewards to improve the reinforcement learning of large language models, achieving better performance with less data compared to traditional methods.
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.
Jingcheng Hu, Yinmin Zhang, Qi Han +3 authors
Open-Reasoner-Zero achieves superior performance on reasoning benchmarks using a minimalist PPO approach, requiring fewer training steps than DeepSeek-R1-Zero.
Zhongyang Li, Ziyue Li, Tianyi Zhou
A novel test-time optimization method, C3PO, improves the accuracy and efficiency of Mixture-of-Experts Large Language Models by re-weighting experts based on surrogate objectives from neighboring samples.
Woongyeong Yeo, Kangsan Kim, Soyeong Jeong +2 authors
UniversalRAG enhances Retrieval-Augmented Generation by dynamically selecting and retrieving knowledge from various heterogeneous and modality-specific sources to improve factual accuracy across different types of queries.
Zhiheng Liu, Ka Leong Cheng, Xi Chen +7 authors
MangaNinjia, a diffusion model-based approach, enhances line art colorization with patch shuffling and point-driven control for precise detail and color matching.
LG AI Research, Kyunghoon Bae, Eunbi Choi +38 authors
EXAONE 4.0 integrates non-reasoning and reasoning modes, supports agentic tool use, and offers multilingual capabilities, demonstrating superior performance in its class.
Zhengxi Lu, Yuxiang Chai, Yaxuan Guo +5 authors
Rule-based reinforcement learning enhances multimodal large language models for GUI action prediction, improving data efficiency and performance.
Stefan Baack, Stella Biderman, Kasia Odrozek +36 authors
Many AI companies are training their large language models (LLMs) on data without the permission of the copyright owners. The permissibility of doing so varies by jurisdiction: in countries like the EU and Japan, this is allowed under certain restrictions, while in the United States, the legal landscape is more ambiguous. Regardless of the legal status, concerns from creative producers have led to several high-profile copyright lawsuits, and the threat of litigation is commonly cited as a reason for the recent trend towards minimizing the information shared about training datasets by both corporate and public interest actors. This trend in limiting data information causes harm by hindering transparency, accountability, and innovation in the broader ecosystem by denying researchers, auditors, and impacted individuals access to the information needed to understand AI models. While this could be mitigated by training language models on open access and public domain data, at the time of writing, there are no such models (trained at a meaningful scale) due to the substantial technical and sociological challenges in assembling the necessary corpus. These challenges include incomplete and unreliable metadata, the cost and complexity of digitizing physical records, and the diverse set of legal and technical skills required to ensure relevance and responsibility in a quickly changing landscape. Building towards a future where AI systems can be trained on openly licensed data that is responsibly curated and governed requires collaboration across legal, technical, and policy domains, along with investments in metadata standards, digitization, and fostering a culture of openness.
Chris Yuhao Liu, Liang Zeng, Yuzhen Xiao +9 authors
Skywork-Reward-V2, a suite of reward models trained on a large-scale, high-quality preference dataset, achieves state-of-the-art performance across various benchmarks by leveraging human-AI synergistic curation.
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