I don't like the prototype in this case either, it heralds too many limitations in use. The practical ones of artificially limiting the number of arrays it can operate on, but more fundementally, the fact that you can only use real arrays.

With the caveat of requiring the user to pass an AoA ref, the function becomes much more flexible and makes writing transpose LOL trivial:

#! perl -slw
use strict;
use Data::Dumper::SLC;

sub multimap (&$) {
    my( $code, $aref ) = @_;
    return unless ref $aref eq 'ARRAY';
    map {
        my $i = $_;
        $code->( map{ $_->[ $i ] } @$aref )
    } 0 .. $#{ $aref->[ 0 ] };
}
my @l = 'a' .. 'h';
my @u = 'A' .. 'H';
my @n = 1 .. 8;

print for multimap { join ', ', @_ } [ \( @u, @n, @l ) ];

sub transpose {
    return multimap{ [ @_ ] } $_[ 0 ];
}

my @LoL = ([1,2,3],[2,3,5]);
my @transposed = transpose \@LoL;

Dump \@transposed;

__END__
P:\test>464573
A, 1, a
B, 2, b
C, 3, c
D, 4, d
E, 5, e
F, 6, f
G, 7, g
H, 8, h
[ [ '1', '2',  ], [ '2', '3',  ], [ '3', '5',  ],  ]
[download]

This is nice, ++ to follow

first, a suggestion - why don't you put [] around your end result map, so we can easily chain multimaps.
there's actually a lot to be said on the topic - for example, 'multimap' may not be the best name for that function.. while 'map' is List -> List ( with proviso on iterators here ), and 'reduce' is List -> Scalar, this new thing is List x List -> List.. It agrees with 'map' on the end List, but it actually provides the dimensionality reduction of 'reduce'..

now, of course, why stop here - one can imagine some sort of morphing operator multireduce(n,m) ListOfBlocks, AoAoA..A that colapses N-dim structures to M-dim ones..

but this goes into academic grounds - I can't imagine this being useful, or easily mastered thing..

lastly, what is Data::Dumper::SLC? I can't find it anywhere.. personal extension ?