Reciprocal rank fusion uses a result's position in each input list to assign credit. A result ranked high in several lists can rise in the combined ordering, even if the search engines score documents on different scales.
For example, a keyword search and a semantic search can each return ten results. Rank fusion can combine the two lists using their positions without comparing their original scores directly.
ELI5
Imagine two librarians each making a list of books for you. A book near the top of both lists earns more points than one near the bottom of a single list.
Adding those points gives you a combined list. Reciprocal rank fusion follows that idea for search results.
Does rank fusion need comparable raw search scores?
It uses rank positions, so the input searches can use different score scales.
What happens if a result appears on only one list?
It can still appear in the combined ordering with credit from that one list.

