Thai, Chinese, and Japanese all have a similar issue.

The tokenizer, the component in full text search that reduces words into searchable segments, (e.g for English, iteration is reduced to iter, so searches for iteration, iterations, iterate, all match), doesn't work well.

The tokenizer breaks words down into individual characters. Searching with "", around the characters helps some. In this case the characters being searched for have to appear near one another and in order. So for the above,

Here's what I saw.
I created WIs with one character
模
块
测
试

The ones with the entries above
模块 测试
模块测试
块模试测
试样(sample)
块状(block)
模拟(simulate)
测绘(measure)

Single character searches always matched any word containing them. So a search for 模 matched 模, 模块, 模块测试, etc...
Multiword searches matched words but didn't care about order of letters, or spaces. A searcgh for 模块 matched  块模试测, 模块测试, etc...
This would be expected given the single character tokens, and a matching algorithm not sensitive to order.

I found slightly better behavior when I put "" (e.g. "模块测试") _around words. Then order mattered, but spaces didn't.
This is because the space character is dropped during matching. Only the tokens and the order matter. The search for "模块测试" didn't match  块模试测, but did match
模块 测试.