思维怪怪 (@0xLogicrw) “DeepSeek 资深研究员陈德里开源了个人项目 Deli AutoResearch SKILL，并发布了由智能” — TopicDigg
思维怪怪
@0xLogicrw
写关于 AI 的一切 Shit Post
AI 信息流：
加入 May 2018
2.4K 正在关注 6.1K 粉丝
DeepSeek 资深研究员陈德里开源了个人项目 Deli AutoResearch SKILL，并发布了由智能体完全自主撰写的第四篇综述论文。
项目以单一的 SKILL.md 协议文件形态呈现，本身不含可执行代码。协议通过规约长周期任务中的状态持久化、防死循环（Anti-Loop）与心跳守护进程（Heartbeat Watchdog），指导 AI 智能体调用子智能体（Subagent）与多角色模拟机制，实现科研全流程的全自动协作。
同时发表的自我博弈综述论文展示了智能体在实验阶段的突破：在零人类干预下，智能体首次自主规划了 GPU 实验，并在 285B 参数量的 DeepSeek 模型上运行强化学习（RL）训练任务。通过 GRPO 算法，智能体完成了从实验设计、编写代码，到运行调试和总结结论的完整研究闭环，并在模拟同行评审中跑出了 8.6/10 的高分。
此前，陈德里已利用同样的方法自主产出过三篇学术综述。首篇关于自主科研智能体的论文经历了约 60 轮智能体迭代，总耗时约 10 小时，实现了从 V1 到 V5 的迭代演进。整套工作流不仅降低了长周期研究的人为操作成本，也验证了 AI 原生研究路径的可行性。
显示更多
🧵 Deli AutoResearch SKILL is now officially open source! 🎉
Alongside it, we’re dropping our 4th survey paper — this time on Self-play.
Inspired by AlphaZero, we got a powerful insight: prior knowledge doesn’t always lift the ceiling.
Models can discover more globally optimal solutions just by playing against themselves.
The biggest change in this paper?
For the first time, the AutoResearch Agent autonomously planned GPU experiments — and submitted actual RL runs on the DeepSeek 285B model.
The entire RL pipeline — experiment design, code writing, running, debugging, and conclusion summarization — was 100% automated, with zero human intervention from me.
This was incredibly difficult, but an incredibly important step.
GRPO is the tool being called by the AutoResearch Agent here.
We see this as the beginning of our Continual Learning research journey. 🚀
As always, this is my personal research project, unaffiliated with any organization. All views are my own.
#AI# #ReinforcementLearning# #SelfPlay# #OpenSource# #AutoML# #ContinualLearning# #DeepSeek#
显示更多
