Pyuyi

@sheriyuo perhaps mainly because of the overflow of successology……

Thank you for reaching out. I genuinely appreciate that people are willing to have these conversations with me. To be honest, I think research is inherently a process filled with uncertainty and stagnation. I remember hearing a quote from Kaiming He that "95% of research is frustration," and I think there is a lot of truth in that. My own life is also full of uncertainty. I do not really know what my future will look like. Should I keep working? Should I apply for a PhD? Should I do both at different stages? On top of that, there is always peer pressure. Seeing new papers every day, watching others publish at an incredible pace, talking to people during interviews, hearing about offers and career trajectories. All of these things can create a lot of pressure. That said, I do not think there is anything wrong with your research journey, or mine, or anyone else's. Research is fundamentally about building things from scratch and exploring the unknown. If everything were already clear and solved, we would just be following instructions. The reason many of us chose research in the first place is because we enjoy exploring questions that nobody has answered yet. Personally, I am drawn to ideas that feel interesting, novel, and underexplored. As long as they are somewhere within the broader LLM space, I am usually happy to spend time thinking about them because I genuinely enjoy the field. As for uncertainty and peer pressure, I do not think they ever completely disappear. When you see someone getting a huge offer or publishing many papers in a year, it is difficult not to compare yourself. But in a strange way, that pressure also gives me a sense of direction. I do not see myself as particularly fast, talented, or resource rich compared to many people around me. The only thing I know how to do is move forward step by step. Read papers. Discuss ideas with others. Write a paper. Submit it. Survive rebuttal. If it gets rejected, revise it and submit again. Repeat. The PhD style of life is often much more boring, repetitive, and exhausting than people imagine. Many days you struggle to find meaning in what you are doing. Sometimes it feels like the goal is simply to graduate and make a living, just like everyone else. Because of that, I think it is important not to let yourself become consumed by frustration. My philosophy has always been fairly simple: work when it is time to work, rest when you are tired, and ask more experienced people for help when you are stuck. No matter what path you choose, you are allowed to pause and recover. Even people doing routine engineering work need that. Research is no different. I wish you all the best, both in research and in life. And if anyone has questions, feel free to DM me. I may not reply immediately, but I am always happy to chat when I have some free time.

@www1110528 谢谢宝(≧∇≦)/

@Pyuyi233 好看羡慕（）

好看喵～

才知道私信要点一下才算开通……

@liuhuavodka 🥰

@Pyuyi233 哇，好漂亮！很有氛围感哦

买了一个投影仪，，发现根本没时间看（） 看会儿星星吧…

@liuhuavodka hh打光的重要性

@Pyuyi233 身材好好

这个光打得……太好看了吧🤤

@sheriyuo 这表看笑了

一眼假，这些带了思考模式的 AI 不可能不是 150 满分，除非拍照的题目 OCR 识别错题干了 之前很多模型厂 1000 收一道能难倒 AI 的数学题（必须给解法+计算量不能过大），找很多 MO 金牌和 Math PhD 出题，现在不从四大上面搬题下来已经难不倒 AI 了（见 Ref） 现在有了 AI 以后，都别说 ICPC / Codeforces 这种竞赛题目了，理科考试题难度完全可以定量衡量了，AI 的平均思考长度就能决定你这个题目的难度，一个很简单的例子，用 Qwen3 thinking mode 去测多少 length 结果饱和： - GSM8K: 512 - MATH500: <3072 - AIME / HMMT: ~32768 更别说是 RL 效果更好、metric 相关性更强的 API 模型了，任何一个测试集都会 saturate 而现在的高考题目根本不配去训练 AI，只能训练人类。一般的竞赛金牌水平也完全打不过 AI 了（参考 ICPC 和某个 IMO exp. 的 Paper 测试），而人觉得的难度已经完全是 AI 觉得的难度了 Ref: 真的造不出来能难住AI的数学题了 bilibili.com/video/BV1sW7y6…

@n2Yp40bIYW91616 腰子…

@Pyuyi233 这是

@wsl8297 山河湖海（）

给老四取个什么名字好呢？

@FascinatingAnon 考古（？

？历化生是要干什么行业的

@SouWu64708 😝

@Pyuyi233 疑似库存😋

累死了…（今天风好大

@henana17va 🥰🥰

@Pyuyi233 宝宝好好看

@Mu_Qiwu 可能是手机优化的好，亦或是我眼睛有点近视，眼睛看到的还不如手机照的😢

@Pyuyi233 好清晰哎

@quasarne 好看喵

@Pyuyi233 好看喵

@z71529284 是的呢，这个还一直在动hh

@Pyuyi233 一直看星星也不错✨

@Saccc_c 111还有吗（？

成都难道真是一个躺平的城市？ 同时在小红书和x发了此贴子，无一人私信 完全不敢想象，要是在深圳约chat, 私信估计都爆了 年轻人还是别来成都创业了

@liuhuavodka rua！

@Pyuyi233 哇 好可爱！揉揉揉揉