I’m learning Polish and it’s kinda tricky because there are so many declensions and conjugations. Lots of words look really similar, and many start with prefixes like w, przez, or po, so it’s harder to recognize them.
On LingQ, every different form of the same word counts as a separate word and in Polish there is a big amount of different forms of the same word. How do you deal with this?
I cannot reply for Polish but for Ukrainian. In Ukrainian you just get used to the declension after having seen them so many times. Now I no more indicate which case is a word because I know it. Even for aspect of a verb with more experience you can recognize them. It’s the same for words which are close but different with a prefix : you get used to them. Same prefix modifies a word in the same way so you may be abble to guess meaning. So be patient. A minimum of knowledge of grammer will greatly help recognizing basic constructions.
With time your word counter will be much higher than for instance in English. 10000 words known in english may be equivalent to 20 000 words in Polish or Ukrainian.
This has been brought up before, and in regards to how to use LingQ in that matter I personally tend to just create LingQ’s for the first few forms of a word I encounter in a lesson and keep the rest, unless there is a premade definition that I can easely takeover. I don’t care much about the stats and only use the unknown word count for identifying the rough difficulty of a lesson. Beeing a bit more forgiving with how many words I create LingQ’s for just results in a slightly higher unknown word count. After a while you will get a feeling for what percentage suits your personal learning needs.
In regards to recognicing the words: I am German so I probably have it a bit easier, as German is somewhat inflected, too. In essence the words are made up the same way all the time, if you ignore irregular forms, and you will familiarize yourself with the LEGO™ stones the language is made of, especially the often reused parts. In general words are often build following the scheme prefix+word stem+suffix, where the word stem describes the semantic, the prefix serves as a semantical modifier and the suffix specifies the word family (noun, verb, et. al.) and the specific grammatical purpose of the word within the sentence (case, tense, et. al.).
While there is a broad variety of word stems, prefixes and suffixes are limited, and after a while you will have memorized the majority of them and thus will start to automatically and rather effortlessly seperate the words into those components, similar to how a kid automatically starts to think of 4∙10+4∙8 when it hears 4∙18 after having practiced multiplication sufficiently long.
The positive thing is, that due to the inflections containing information about the grammatical purpose, those languages rely much less on word order. So even though they usually have some sort of standard word order, one can often purposely break with those without distorting the meaning of the sentence, but adding to its nuances. This is the reason why German for example is often refered to as language of the poets and thinkers, while Mark Twain stated: “If it is to remain as it is, it ought to be gently and reverently set aside among the dead languages, for only the dead have time to learn it.” 
So what exactly is your question? Are you asking how to remember the words or how to mark them?
Words should be marked individually based on word forms (not root forms). In my opinion, with highly inflected languages, if you know the dictionary meaning (nominative form) and grammatical implication (meaning of declined form) both then it should be marked as known. If you know just the dictionary meaning but not really the grammatical implication then you can marked it half-known.
As far as remembering the words, you can probably just memorize the declension tables for a few words just to get the patterns into your head. Then after that it just comes down to brute force reading to increase your total words read. Just don’t bite off more than you can chew by beating your head against the wall with material that has too many unknown words. It’s better to err on the side of reading easy material quickly rather than difficult material slowly.