You would probably find TinyLFU an interesting approach, as it focuses on admission rather than eviction. This is done by using a popularity sketch to probabilistically estimate the frequency of the new arrival compared to the victim, choosing whichever has the highest frequency. You may also find this project's simulator a useful resource, as LIRS is difficult to implement correctly due to subtleties in the paper.
You would probably find TinyLFU an interesting approach, as it focuses on admission rather than eviction. This is done by using a popularity sketch to probabilistically estimate the frequency of the new arrival compared to the victim, choosing whichever has the highest frequency. You may also find this project's simulator a useful resource, as LIRS is difficult to implement correctly due to subtleties in the paper.