Ask HN: Any exciting under-the-radar discoveries/gems from NeurIPS?

So I’m now at a startup and somewhat removed from cutting-edge academic machine learning, but fondly remember the times I attended NIPS, and how many amazing and promising papers/posters introducing brand new techniques and areas of inquiry that were incredibly inspirational, but without the backing of big players and thus not really publicized unless you happened to see them. For those who attended, did any of these stand out? Any good notes people have taken and published?