Ask HN: Interesting recent papers in recommender systems?

I'm trying my hand at implementing a recommender system for mainly tabular data (think song ratings based on tempo, genre, etc.) using the algorithms in Surprise [1] as a baseline.

I assume there are plenty of "throw an LLM at it" papers out there, but are there any interesting architectures/results people have found in "plain" recommender systems recently?

In particular I'd be interested to see how XAI techniques are applied in user-facing recommenders (e.g. "tempo matters a lot to your song ratings")

[1] https://surpriselib.com/