zhiyunyu

MS in Electrical Engineering @UBuffalo | Opinions are my own

[zhiyunyu @Hamanansu]

@sheriyuo brain had at last evolved（大脑升级），，，

[Zhiqiang Shen @szq0214]

🎉🎉🎉 Excited to introduce our recent project FigMirror - a very interesting and useful tool with a simple workflow for making any paper-style figures.

- See a beautiful figure in a paper
- Screenshot it
- Add your own data
- Get a new Matplotlib figure with the same visual style

FigMirror learns the quiet details that make paper figures look polished:

- typography
- spacing
- line weight
- color restraint
- layout rhythm

The key mechanism is Grounded Measurement.

Computer-use AI can point to coordinates inside the reference figure. Code then inspects the pixels, colors, spacing, and layout around those points. This gives the system concrete visual evidence to iterate on.

Our FigMirror draws a candidate figure, compares it with the reference, keeps what works, and improves what still feels off.

Outputs:

- editable Matplotlib code
- camera-ready PDF

It works as both a local Web UI and a Codex / Claude Code skill.

Open source:
https://t.co/zCVM0aAyxp

Try it before your next deadline!

[zhiyunyu @Hamanansu]

哎最近业界好恐怖

[zhiyunyu @Hamanansu]

@chaos_sens @excited_naive 啊怎么会呢 我2011年就开始看动画了 竹达彩奈就是彩喵呀

[Sanbu 散步 @sanbuphy]

花了段时间写了 RL 教程 Hands-On Modern RL，路线是从 CartPole + PPO 入门，然后到 LLM 后训练（RLHF、DPO、GRPO）、Agentic RL。代码先行，公式用来解释现象。英文版很快更新。
目前是草稿版本，RLHF、Agentic RL 部分本地审校中。
欢迎提 PR 或 Issue & 显卡支持：https://t.co/PtgSByY96U

[Yann LeCun @ylecun]

This could have covered the entire budget of the National Science Foundation for 10 years.

Instead, Trump wants to reduce the NSF budget by 50% ($5B a year instead of $9B), which would decimate the American scientific research ecosystem, dramatically reduce the number of PhD graduates, and destroy the technology innovation flywheel.

[zhiyunyu @Hamanansu]

So my own local library stocks gtms now.

[zhiyunyu @Hamanansu]

@sheriyuo 还很年轻啊 恭喜恭喜 祝你研究顺利🙏㊗️🙏🎊🍾

[Xiuyu Li @sheriyuo]

写的非常详细的入门教程

RL 很恶心的点就在于你什么都要会：pretrain、inference、theory、infra

现在我也有了可以推荐给别人入门的读物，比我当时东看一点西看一点好太多了🥺

很多现代 LLM 的教学其实跳过了相当一部分的基础知识，比如 CNN、Q-Learning 这一类经典内容。我感觉我也是填鸭式地在学习这些内容，需要耐下心来慢慢地去看那些最旧最老最传统的东西

很感谢苏剑林的 blog 带我入门，我也想能写出一些带别人入门的东西，只不过即使是用 LLM 辅助写 notes 也相当折磨人，然后知乎上还没什么人看

当然还是有很多人喜欢看我写的垃圾 notes 的，还被行业杰青大佬订阅了，这种事跟 paper 被人关注了一样开心

Website: https://t.co/TR35oZGB7J

[zhiyunyu @Hamanansu]

@MoonLightBiskit 世界上最好的航空公司：WN

[adafruit industries @adafruit]

Espressif unveils the ESP32-S31, a dual-core RISC-V SoC with Wi-Fi 6, Bluetooth 5.4

https://t.co/8Hcq4HRSvh

[Xiuyu Li @sheriyuo]

Getting blocked by stupid KOL like this is honestly an honor.

RL is still a great path. And if you feel your math talent is being wasted, there is a lot of deep work waiting in policy gradients, optimization theory, etc.

It is always strange to see people with unclear credentials lecturing MS and PhD students about their future, as if studying math or physics makes you second class while importing packages or vibe coding is somehow the peak of computer science.

[Xiuyu Li @sheriyuo]

PhD level research is a completely different game from MO, PhO or OI.

I was also one of the many people who "failed" at OI, but that never stopped me from doing well in regular academics comparably. It also made me realize that I was never that interested in qr/qd for its own sake.

In some sense, the current research environment is naturally much more ADHD friendly than school or standardized testing. Hyperfixation, jumping across ideas, working on many agentic workflows, and chasing unusual connections can actually become an advantage if you learn how to direct it.

What I actually enjoy is math itself, and the feeling of doing AI research. Well thanks my supervisor, he found me.

[CNX Software @cnxsoft]

Open-source, self-hosted Arduino, Raspberry Pi, and ESP32 simulator.
https://t.co/8w5RClkdYp

Like the Wokwi project, Velxio simulates popular development boards right in your web browser. But the difference is that the open-source project can be self-hosted, running on your own machine.

The project currently supports 19 boards and 48 components. Other highlights include support for multiple boards (e.g., Arduino connected to ESP32) and full QEMU emulation support for ESP32 and Raspberry Pi 3 (Linux).

[[email protected] @Jiaxi_Cui]

帮转：在 X 上被大吹特吹的 Google TurboQuant，实际上剽窃了我朋友重要的科研成果 RabitQ

并且发表在今年的 ICLR 上，有充足证据证明 Google 团队知情并且隐瞒了他们先前论文的成果和贡献

因为我遇到过 paper、idea 被同组师兄剽窃的经历，三年过去了，那篇 paper 至今还是他们 Google Scholar上引用最多的成果

虽然我依旧保留着证据等合适时间给他们致命一击出口恶气。但当时没有任何人为我发声，导师只说了一句对不起，这种学术不端的行为让人非常反感

https://t.co/khxG2iBjKQ

[zhiyunyu @Hamanansu]

你们能怎么办只有天知道，当然发生了什么事，我可以做到两眼只当没看到，我已经润了。

[zhiyunyu @Hamanansu]

我觉得我们00后已经可以有盖棺定论了，我们是进入大学后父代创伤集体性爆发的一代，也是即将被就业和经济现实牺牲的一代，这种能量会在未来再爆发一次。
但究竟是正面的还是负面的，我个人倾向是负面的，因为“权力层面上的负责，accountable”这一代的权责对等义务的意识天生就是没有。

[zhiyunyu @Hamanansu]

@magicxqq Powered by Kawasaki

[zhiyunyu @Hamanansu]

https://t.co/zU54tkQFfa

[zhiyunyu @Hamanansu]

https://t.co/MbWfOrxTJR

[zhiyunyu @Hamanansu]

https://t.co/xgMIXXMNrn