Hi, my name is Ian! I'm a recent graduate from UC Irvine with a B.S. in Business Information Management. I am passionate about blending business and technology, in which I have gained demonstrated expertise across both fields.
Outside of work, I really enjoy music! The widget below shows what I'm currently listening to on Spotify, so hit me up if you have any recs!
Not currently playing on Spotify
Probes 15 open-weight models (0.27B-70B params) across four model families and finds that their ability to tell evaluation apart from real deployment follows a clear power-law with model size, letting researchers forecast this behavior in future, larger models.
Shows that cutting reasoning models off mid-chain-of-thought can hurt them more than giving no reasoning at all, and that how a model reasons (code vs. natural language) changes how gracefully it degrades under tight token budgets across four frontier models.
Feel free to reach out to me via email!
Say Hello