Yu Yang (@yang_ml_estate) 's Twitter Profile
Yu Yang

@yang_ml_estate

PhD, PE, PMP. Data scientist. @MIZZOU Tiger. Interested in LLM, AI, Webapp. Side Hustle = Real Estate. Opinions = mine.

ID: 1721569087887724544

calendar_today06-11-2023 16:44:22

1,1K Tweet

644 Followers

529 Following

Yu Yang (@yang_ml_estate) 's Twitter Profile Photo

Welcome back to the LLM and Reinforcement Learning podcast. This is the very first episode in our GRPO series, where our goal is to decode the Group Relative Policy Optimization algorithm piece by piece. We’ll start today with the foundation: Kullback–Leibler divergence. From the

Welcome back to the LLM and Reinforcement Learning podcast. This is the very first episode in our GRPO series, where our goal is to decode the Group Relative Policy Optimization algorithm piece by piece. We’ll start today with the foundation: Kullback–Leibler divergence. From the
Yu Yang (@yang_ml_estate) 's Twitter Profile Photo

PMF should be MPF: Market Product Fit. Market should be at the first place. Next question: where to find the niche market? Any idea?

Yu Yang (@yang_ml_estate) 's Twitter Profile Photo

Claude still could not resolve the issue on rendering equations after two years… Are they giving up all app users?🤣🤣 Apparently OpenAI does a much better job on it .

Claude still could not resolve the issue on rendering equations after two years…

Are they giving up all app users?🤣🤣

Apparently OpenAI does a much better job on it .
Yu Yang (@yang_ml_estate) 's Twitter Profile Photo

Practicing leetcode is still needed from time to time, even we are vibe coding with Claude Code or Codex. What do you think?

Yu Yang (@yang_ml_estate) 's Twitter Profile Photo

Started listening to Chinese podcasts on Xiaoyuzhou since March 2025— mostly on LLMs, VLMs, AI product building, and AI organization management. Logged ~142 hours so far. It’s the first time Chinese has directly boosted my professional work in AI 😁 Here are episodes I had

Started listening to Chinese podcasts on Xiaoyuzhou since March 2025— mostly on LLMs, VLMs, AI product building, and AI organization management. Logged ~142 hours so far. It’s the first time Chinese has directly boosted my professional work in AI 😁

Here are episodes I had
Yu Yang (@yang_ml_estate) 's Twitter Profile Photo

After working on it off and on for over three months, I finally finished Part 3 of my DeepSeek GRPO series on medium. If you dive deep to DeepSeek R1 and DeepSeek Math, you’ll see exactly how LLM reasoning improves, where the ceiling is, and why high-quality labeling still

After working on it off and on for over three months, I finally finished Part 3 of my DeepSeek GRPO series on medium.

If you dive deep to DeepSeek R1 and DeepSeek Math, you’ll see exactly how LLM reasoning improves, where the ceiling is, and why high-quality labeling still
Yu Yang (@yang_ml_estate) 's Twitter Profile Photo

ChatGPT just saved me $3,000. The bath tub of one rental property is old with some ugly glue. To replace it with a new one, the labor cost will be around $3,000. I just asked the ChatGPT, and got a tip of fixing it with glass fiber. It only costed me $325.

ChatGPT just saved me $3,000. 

The bath tub of one rental property is old with some ugly glue. To replace it with a new one, the labor cost will be around $3,000. I just asked the ChatGPT, and got a tip of fixing it with glass fiber. It only costed me $325.