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。
Chuyan Chen, Haoxing Chen, Kun Chen +27 authors
LLaDA-Image unifies a 6B diffusion transformer with a frozen vision-language module, using image-only pre-training and a Muon optimizer to generate photorealistic images with precise editing, and is distilled into a fast 2-4 step variant that achieves state-of-the-art open-source results.
Yang Yue, Zhiqi Chen, Rui Lu +5 authors
Despite initial claims, reinforcement learning with verifiable rewards does not introduce fundamentally new reasoning abilities to LLMs, instead it enhances performance by biasing output distribution toward rewarded paths without expanding the reasoning boundary.
Guibin Zhang, Leo Lu, Fangzhou Xie +13 authors
JIT-Agent is a trainable model that synthesizes adaptive agent harnesses for off-the-shelf LLMs, improving performance across diverse models and tasks.
Songchun Zhang, Zeyue Xue, Siming Fu +6 authors
Astrolabe is an efficient online reinforcement learning framework for distilled autoregressive video models that improves generation quality through forward-process RL formulation and streaming training with multi-reward objectives.
Zheyuan Yang, Lyuhao Chen, Arman Cohan +1 authors
Two post-training strategies, distillation and RLVR, enable inference-time scaling in table reasoning tasks, resulting in a model (Table-R1-Zero) that matches GPT-4.1's performance using fewer parameters and shows strong generalization.
Fanxu Meng, Zengwei Yao, Muhan Zhang
TransMLA converts GQA-based models to MLA to reduce KV cache size and improve efficiency without sacrificing expressiveness.
Kanana LLM Team, Yunju Bak, Hojin Lee +26 authors
Kanana, a series of bilingual language models, achieves superior performance in Korean and competitive performance in English with lower computational costs through efficient pre-training and post-training techniques.
Sanchit Gandhi, Patrick von Platen, Alexander M. Rush
Distil-Whisper, a smaller and faster variant of the Whisper model, achieves nearly the same performance with fewer resources and is optimized for low-latency environments.
Jinhyuk Lee, Zhuyun Dai, Xiaoqi Ren +17 authors
Gecko is a compact and versatile text embedding model that distills knowledge from large language models to achieve strong retrieval performance on benchmarks.
Junxiong Wang, Daniele Paliotta, Avner May +2 authors
Large-scale Transformer models can be distilled into linear RNNs using attention layer weights, achieving competitive performance with enhanced deployment efficiency.
Xingchao Liu, Xiwen Zhang, Jianzhu Ma +2 authors
Rectified Flow is used to develop an ultra-fast one-step text-to-image generator named InstaFlow, achieving high image quality with significantly reduced inference time compared to existing methods.
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