Exponentiation is right-associative.

[4,3,6] == 4 ^ 3 ^ 6 == 4 ^ (3 ^ 6) == 4 ^ 729 == 7975054838978857957869783595079038148815929134873574372267239147002092943367488761318118323875971614518937350748146290561474995443284105875630084237401365691741820338695355818455664737390057540900825684807418526691288993393970109201605650784885019821709886980841347925777256289930194351608431522864880360956238794237301100891415162717067858756153598763756850431348717233286605102660510814000306243542621599170963477008725517984745855647744

Are these not nested lists?

You're reading the code of tests wrong. What you see there is a list of test cases, where each test case is a tuple of (input_list, expected). If you read tests further, you see:

for test_input, test_output in test_data:
    test.assert_equals(last_digit(test_input), test_output)

You can see that the call looks like: last_digit(test_input), where test_input is a single, flat list. One list from the whole test set, one by one.

The above picture you posted is not at all nested lists. More like tuples that have lists in them.

What do you mean, "arguments that hare nested lists"? There should be no nested lists anywhere. Why do you think that you are given nested lists as inputs?

See this fork with an explanation.