Ask HN: How do you come with research ideas? <focus on LLMs>

[toombowoombo]

Hi HN-ers,

I'm looking for research ideas for my master thesis and I'm quite bad at determining what's feasible to do within 6-10 worth of work (approx. 15h/week).

Currently I'm thinking about either measuring bias/variance in LLMs or looking into time-series applications of transformer-based models, but I'm still in the exploration phase and I'm very much open to any thoughts/remarks/suggestions :DD

Thanks & wishing y'all a nice end of February & nicer beginning of March ^^