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。
NeoHorse Team, Guoliang Cao, Guohao Dai +34 authors
NeoHorse-1 uses agentic post-training with intelligent routing, structured feedback loops, and curriculum-based distillation to improve model capabilities across agent benchmarks.
Mingyu Ouyang, Yiran Wang, Siyuan Hu +9 authors
Player-1 improves game control by combining supervised fine-tuning on gameplay trajectories with reinforcement learning across synthetic environments, achieving substantial gains over demonstration-only baselines.
Jinguo Zhu, Weiyun Wang, Zhe Chen +44 authors
InternVL3 is a multimodal pre-trained language model that jointly learns from both multimodal data and text, improving performance and scalability through advanced techniques and setting a new state-of-the-art in multimodal tasks.
Jian Yang, Xianglong Liu, Weifeng Lv +68 authors
A comprehensive guide to code LLMs, covering their lifecycle from data curation to deployment, including techniques, trade-offs, and research-practice gaps.
Hao Liang, Qihan Lin, Meiyi Qiang +5 authors
OmniEdu is a family of instruction-tuned educational foundation models that improve K-12 problem solving, curriculum understanding, and tutoring through capability-balanced training data.
Jie Wu, Zhenru Zhang, Beichen Zhang +11 authors
Terminal-Universe reconstructs executable workspaces from agent trajectories to synthesize diverse training tasks and improves post-training performance through supervised fine-tuning.
Tianxin Wei, Ting-Wei Li, Zhining Liu +26 authors
Agentic reasoning redefines large language models as autonomous agents capable of planning, acting, and learning through continuous interaction in dynamic environments across single-agent and multi-agent frameworks.
Julianna Piskorz, Antonin Berthon, Mihaela van der Schaar
In controlled distillation experiments, rollout policy has limited impact compared to token-level KL direction and learning rate, with on-policy benefits depending on objective and hyperparameters.
Yongliang Wu, Yizhou Zhou, Zhou Ziheng +7 authors
Dynamic Fine-Tuning (DFT) improves the generalization of Large Language Models (LLMs) by dynamically rescaling gradients, outperforming standard Supervised Fine-Tuning (SFT) and showing competitive results in offline reinforcement learning.
Junlin Yang, Che Jiang, Yu Fu +21 authors
OpenMLE enables recursive self-improvement in machine learning engineering via a full-stack system combining execution-grounded operator learning and long-horizon evolutionary search, significantly boosting benchmark performance with a 35B model.
Yafu Li, Runzhe Zhan, Haoran Zhang +25 authors
A systematic approach transforms post-trained reasoning models into rigorous olympiad-level solvers through reverse-perplexity curriculum, two-stage reinforcement learning, and test-time scaling, achieving gold-medal performance on mathematical and physics competitions.
DeepCybo Team, Yu Bin, Haipeng Cao +51 authors
PhysBrain 1.5 unifies physical environment understanding, action generation, and future state prediction via joint autoregressive training on discrete vision-language, motion, and visual target sequences, achieving state-of-the-art open-source embodied performance.
Qize Yu, Lianrui Fan, Boyu Chen +23 authors
GroundingPI is a 4B grounding foundation model that improves precise object localization and downstream robotic and autonomous driving performance through multimodal pretraining and reinforcement learning.
Shun Zou, Ziyu Ma, Yi Zou +6 authors
TerminalHorizon automates long-horizon terminal agent data synthesis by reconstructing environments around omitted work and progressively expanding tasks, yielding substantial gains across coding, production, and reasoning benchmarks.
Jiyuan Wang, Chunyu Lin, Lei Sun +8 authors
RL3DEdit uses reinforcement learning with rewards from a 3D foundation model to achieve multi-view consistent 3D editing from 2D editing priors.
Qiutong Chen, Yuchan Guo, Zhenlong Yuan +10 authors
VepAgent improves video event prediction by combining causal reasoning with tool-augmented reinforcement learning and a diagnostic tool library to recover missing visual evidence.
Aviral Kumar, Vincent Zhuang, Rishabh Agarwal +15 authors
The multi-turn online reinforcement learning approach SCoRe improves LLM self-correction using self-generated data and regularization, achieving state-of-the-art performance.
Siming Huang, Tianhao Cheng, Jason Klein Liu +16 authors
OpenCoder is a top-tier open-access code LLM that provides comprehensive reproducible resources, including data, pipelines, and protocols, to advance code AI research.
Shen Nie, Fengqi Zhu, Zebin You +7 authors
LLaDA, a diffusion model trained from scratch, outperforms autoregressive models in benchmarks and demonstrates strong instruction-following capabilities, challenging the dominance of ARMs in LLMs.
Zongjie Li, Alan Z. W, John Nicolas J +4 authors
A data-centric framework with specialized systems for reasoning analysis, cost reduction, and execution verification enables small teams to train open-weight cyber agents that achieve top-tier performance on benchmark suites.
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