Menu
SkillsMP has collected 3 skills from XiaoRed5/Agentic-RL-Most-Detailed-Intro. Open a skill to review its source and details.
Showing 3 of 3 collected skills.
Research and explain a paper, algorithm, training recipe, data-construction method, or code mechanism in plain language, then write one ready-to-paste Chinese prompt for Qwen image-generation models to draw an accurate mechanism diagram. Use when the user…
Source text: Chinese
先理解用户问题并定向研究论文、附录与代码，还原方法的具体实现，再把机制转译成信息密度可调、图形优先的教学图；支持确定性 SVG、ImageGen 构图探索与混合矢量化。用于“讲清论文方法怎么实现”“画机制图”“可视化算法”“把公式或训练流程画出来”“解释 code mid-training、数据构造、监督信号、训练与推理差异”等任务。区别于实验结果图 paper-figure、普通流程图 mermaid-diagram 和图像描述 figure-description。
Source text: Chinese
Automate the full AutoDL GPU experiment lifecycle for Agentic RL/LLM projects: choose and rent a GPU within an approved budget, package and upload code, download Hugging Face or ModelScope models to the data disk, create environments, preflight…