Trends · 研究趋势
数据来自 Hugging Face 论文的 AI 提取关键词,按月统计研究方向的增长与热度。
Kimi Team, Tongtong Bai, Yifan Bai +399 authors
Kimi K3 is a large-scale mixture-of-experts model with native vision and long-context capabilities that improves scaling efficiency and achieves strong performance across coding, reasoning, and agentic tasks.
Qwen, An Yang, Baosong Yang +39 authors
Qwen2.5, an enhanced series of large language models, demonstrates superior performance across various benchmarks and use cases through extensive pre-training and advanced post-training techniques.
Chujie Zheng, Shixuan Liu, Mingze Li +9 authors
Group Sequence Policy Optimization (GSPO) is a reinforcement learning algorithm that improves training efficiency and performance of large language models by using sequence-level importance ratios and operations.
Haifeng Wang, Hua Wu, Tian Wu +435 authors
ERNIE 5.0 is a production-scale trillion-parameter autoregressive model that unifies multimodal understanding and generation through sparse MoE architecture and elastic training.
Changhai Zhou, Kieran Liu, Yuhua Zhou +17 authors
LongStraw is an execution framework that enables million-token reinforcement learning post-training with fixed GPU resources by reducing live training graphs through replay and selective state retention.
GLM-4. 5 Team, Aohan Zeng, Xin Lv +168 authors
GLM-4.5, a Mixture-of-Experts large language model with 355B parameters, achieves strong performance across agentic, reasoning, and coding tasks using multi-stage training and reinforcement learning.
Ailin Huang, Ang Li, Aobo Kong +212 authors
Step 3.5 Flash is a sparse Mixture-of-Experts model that achieves frontier-level agentic intelligence through efficient parameter utilization and optimized attention mechanisms, demonstrating strong performance across multiple benchmarks.
Haiwen Diao, Penghao Wu, Hanming Deng +55 authors
Unified vision-language models treat understanding and generation as integrated processes rather than separate tasks, demonstrating strong performance across multiple multimodal capabilities including image synthesis and action reasoning.
Kwai Keye Team, Bin Wen, Changyi Liu +50 authors
Kwai Keye-VL-2.0-30B-A3B is an open-source Mixture-of-Experts multimodal foundation model that enables long-video understanding and agentic intelligence through DeepSeek Sparse Attention and specialized training infrastructure.
Meituan LongCat Team, Anchun Gui, Bei Li +159 authors
A 560-billion-parameter Mixture-of-Experts reasoning model achieves state-of-the-art performance on agentic benchmarks through a unified training framework combining domain-parallel expert training with fusion, along with enhancements for real-world robustness and complex reasoning.
An Yang, Baosong Yang, Binyuan Hui +55 authors
The Qwen2 series, comprising 0.5 to 72 billion parameter models, surpasses prior open models across language understanding, generation, multilingualism, coding, math, and reasoning, with exceptional performance in benchmarks like MMLU, GPQA, HumanEval, GSM8K, BBH, MT-Bench, Arena-Hard, and LiveCodeBench.
Shuai Bai, Yuxuan Cai, Ruizhe Chen +61 authors
Qwen3-VL, a vision-language model, excels in text and multimodal understanding through advanced architectures and larger contexts, achieving superior performance across benchmarks.
SII-OpenMOSS Team, Donghua Yu, Mingshu Chen +37 authors
MOVA is an open-source model that generates synchronized audio-visual content using a Mixture-of-Experts architecture with 32 billion parameters, supporting image-text to video-audio generation tasks.
Shuqi Lu, Chaofan Li, Kun Luo +21 authors
AREX is a recursively self-improving deep research agent that verifies answers constraint-wise, compresses verified evidence into a compact state, and uses targeted follow-up research to refine results over long horizons.
Brandon McKinzie, Zhe Gan, Jean-Philippe Fauconnier +28 authors
A study of Multimodal Large Language Models finds a mix of pre-training data and image encoder design crucial for SOTA performance on multimodal benchmarks.
Chujie Zheng, Kai Dang, Bowen Yu +7 authors
The paper provides a theoretical foundation for optimizing sequence-level rewards in reinforcement learning using token-level objectives, highlighting the importance of techniques like importance sampling correction, clipping, and Routing Replay for stabilizing training, especially with large language models.
Shengbang Tong, David Fan, John Nguyen +18 authors
Controlled multimodal pretraining experiments reveal key insights about unified visual representations, data complementarity, world modeling emergence, and efficient scaling through mixture-of-experts architectures.
Hong Liu, Jiaqi Zhang, Chao Wang +13 authors
Embedding scaling offers superior sparsity scaling compared to expert scaling in large language models, enabling efficient inference through system optimizations and speculative decoding.
Lei Bai, Zongsheng Cao, Yang Chen +47 authors
Agents-A1, a 35B Mixture-of-Experts Agentic Model, achieves trillion-parameter-level performance through long-horizon trajectory scaling and heterogeneous agent ability scaling via a three-stage training approach involving supervised fine-tuning, domain-level teacher models, and multi-teacher distillation.
Shaowen Wang, Ge Zhang, Kairong Luo +6 authors
Looping middle layers in sparse Mixture-of-Experts Transformers improves training efficiency and downstream performance while matching per-token FLOPs, parameters, and cache budgets.
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