>Find a posttraining paper
>Ask if it's proper RL or just weighted BC
>They don't understand
>Pull out Sutton&Barto and explain what is BC and what is proper RL
>"It's a good approach, sir"
>Read the paper
>It's BC
we introduce a new RL technique which trades off variance for bias, which has the benefit of lower variance. this has higher stability when compared to bias-free approaches for small runs with tiny batch size, which kept crashing for us. further scaling analysis is needed
It must be really frustrating in announcements to see "Mila / MIT / ETH ... etc. researchers discovered X" or when everything is attributed to the most senior researcher like "Hinton / Sutton paper did X" ... lead authors have names and its crucial that we *say their names*.
@Laz4rz Homegirl has 7k citations and an h-index of 17.
That's pretty insane for a grad student, definitely not an average experience even for PhDs from top schools