Mova0.5 pre-release⌕

☰

home / corpora / questions / AI / Hallucinations: confident made-up answ… / t4

Test 4 · ★ · Hallucinations: confident made-up answers

In a cartoon, a bot gives the exact same answer every single time, forever. What does the cartoon get wrong?

Mova's choice: option 1 · The right answer: option 1 · right ✓ · Who acts: bot, cartoon · What happens: give, get · To what, where: exact same answer, cartoon

This is test 4 · ★ of lesson 5 «Hallucinations: confident made-up answers» of the AI course of Mova's primary school. The page gives the question's syntax tree — a dependency tree in the style of Universal Dependencies — and its pragmatics clause by clause, then the 4 options as the student saw them, each with its tree and pragmatics. Every text is parsed by Mova itself. How the student answered it, with the score of every option and why: the test at school.

The question

Sentence 1: In a cartoon, a bot gives the exact same answer every single time, forever.

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	In	in	ADP	_	3 cartoon	case
2	a	a	DET	Definite=Ind PronType=Art	3 cartoon	det
3	cartoon	cartoon	NOUN	Number=Sing	7 gives	obl
4	,	,	PUNCT	_	3 cartoon	punct
5	a	a	DET	Definite=Ind PronType=Art	6 bot	det
6	bot	bot	NOUN	Number=Sing	7 gives	nsubj
7	gives	give	VERB	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	root	root
8	the	the	DET	Definite=Def PronType=Art	11 answer	det
9	exact	exact	ADJ	Degree=Pos	11 answer	amod
10	same	same	ADJ	Degree=Pos	11 answer	amod
11	answer	answer	NOUN	Number=Sing	7 gives	obj
12	every	every	DET	PronType=Tot	14 time	det
13	single	single	ADJ	Degree=Pos	14 time	amod
14	time	time	NOUN	Number=Sing	7 gives	obj
15	,	,	PUNCT	_	16 forever	punct
16	forever	forever	ADV	_	7 gives	advmod
17	.	.	PUNCT	_	7 gives	punct

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	give	present	a bot	the exact same answer		a cartoon	time	forever property	vn give-13.1-1 «NP V NP»: agent=Agent answer=Theme

One row per clause the reader finds. «not said here»: a role filled from the clause above. «roles by»: a VerbNet class and frame when one fits the verb, else the reader's rules on the tree.

Sentence 2: What does the cartoon get wrong?

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	What	what	PRON	PronType=Int	5 get	obj
2	does	do	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	5 get	aux
3	the	the	DET	Definite=Def PronType=Art	4 cartoon	det
4	cartoon	cartoon	NOUN	Number=Sing	5 get	nsubj
5	get	get	VERB	Mood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Fin	root	root
6	wrong	wrong	ADJ	Degree=Pos	5 get	xcomp
7	?	?	PUNCT	_	5 get	punct

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	get	present · do-support	the cartoon					wrong property	vn comprehend-87.2-1 «NP V NP»: agent=Experiencer what=Stimulus

The options

Option 1 · ← Mova's choice · ✓ the right answer

Real bots add randomness and can answer differently

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	Real	real	ADJ	Degree=Pos	2 bots	amod
2	bots	bot	NOUN	Number=Plur	3 add	nsubj
3	add	add	VERB	Mood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Fin	root	root
4	randomness	randomness	NOUN	Number=Sing	3 add	obj
5	and	and	CCONJ	_	7 answer	cc
6	can	can	AUX	VerbForm=Fin	7 answer	aux
7	answer	answer	VERB	VerbForm=Inf	3 add	conj
8	differently	differently	ADV	_	7 answer	advmod

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	add speech: assert	present	Real bots	randomness					vn mix-22.1-2 «NP V NP»: agent=Agent randomness=Patient
2	answer speech: assert	can	Real bots not said here					differently property	vn reciprocate-112-1 «NP V»: agent=Agent

Option 2

Nothing, that's exactly how bots work

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	Nothing	nothing	PRON	Number=Sing PronType=Neg	root	root
2	,	,	PUNCT	_	1 Nothing	punct
3	that	that	PRON	Number=Sing PronType=Dem	8 work	nsubj
4	's	be	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	8 work	cop
5	exactly	exactly	ADV	_	8 work	advmod
6	how	how	ADV	PronType=Int	8 work	advmod
7	bots	bot	NOUN	Number=Plur	8 work	nsubj
8	work	work	VERB	Mood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Fin	1 Nothing	acl:relcl

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	is	present						exactly, work property	tree rules

Option 3

Real bots always answer twice

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	Real	real	ADJ	Degree=Pos	2 bots	amod
2	bots	bot	NOUN	Number=Plur	4 answer	nsubj
3	always	always	ADV	PronType=Tot	4 answer	advmod
4	answer	answer	VERB	Mood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Fin	root	root
5	twice	twice	ADV	NumForm=Word NumType=Mult	4 answer	advmod

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	answer speech: assert	present	Real bots					always, twice property	vn reciprocate-112-1 «NP V»: agent=Agent

Option 4

Real bots never answer at all

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	Real	real	ADJ	Degree=Pos	2 bots	amod
2	bots	bot	NOUN	Number=Plur	4 answer	nsubj
3	never	never	ADV	PronType=Neg	4 answer	advmod
4	answer	answer	VERB	Mood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Fin	root	root
5	at	at	ADP	_	6 all	case
6	all	all	DET	PronType=Tot	4 answer	obl

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	not answer speech: assert	present	Real bots			all			vn reciprocate-112-1 «NP V»: agent=Agent

Text: the tests of Mova's primary school, written for Mova (world/data/school/), as the course runner gave them to the student. Trees: Mova's own parse — the en model on the CPU, when the site is built — not checked by a teacher; they may be wrong. Frames: Mova's frames reader (prag), each text read on its own. Elamur Cosmic License.

This page in Markdown → https://md.mova.elamur.ai/free/en/corpora/questions/ai/5-hallucinations-confident-made-up-answers/t4/