Jev has blind spots.
Jev's memory fades on obscure facts. It gets 73% of the questions about the most viewed quarter of songs on Wikipedia right, and 49% about the least viewed. Tricky wording trips it. It gets 70% of the plain control questions and 50% of the word traps. Two hops are hard. One kind of question asks whether a song came out the same month as another event. On 26 of them, Jev knew both facts on their own. It got the chained question right on only 13.
Nothin' Shakin' is an obscure song, and George sings it.
The numbers come from the bench's record for this slide.
Run it
john.is_a_song="The text is the title of a song by the Beatles."
question="$is_a_song Who sings the lead vocal on it?"
singers=(
--option "john=John Lennon"
--option "paul=Paul McCartney"
--option "george=George Harrison"
--option "ringo=Ringo Starr"
)
printf '%s' "Nothin' Shakin'" |
thinkthen choose "$question" "${singers[@]}" \
--replay recording"john"
Change the bar
is_a_song="The text is the title of a song by the Beatles."
question="$is_a_song Who sings the lead vocal on it?"
singers=(
--option "john=John Lennon"
--option "paul=Paul McCartney"
--option "george=George Harrison"
--option "ringo=Ringo Starr"
)
printf '%s' "Nothin' Shakin'" |
thinkthen choose "$question" "${singers[@]}" \
--threshold 0.5 \
--replay recordingnull
The lesson
With no bar, the answer is John, and John is wrong. A bar of 0.5 sends this question to a person.
A low probability is Jev saying it does not know.
On GitHub: github.com/botassembly/beatles-bench