Mova0.5 pre-release⌕
☰
home / corpora / questions / AI / Hallucinations: confident made-up answ… / t5

Test 5 · ★ · Hallucinations: confident made-up answers

Two bots disagree about how tall a mountain is. What does that tell you?

Mova's choice: option 1 · The right answer: option 1 · right ✓ · Who acts: Two bots, mountain, that · What happens: disagree, be, tell · To what, where: you · Speech: someone speaks here

This is test 5 · ★ 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: Two bots disagree about how tall a mountain is.

nummodnsubjcaseadvmodobldetnsubjcoppunctrootTwoNUMbotsNOUNdisagreeVERBaboutADPhowADVtallADJaDETmountainNOUNisAUX.PUNCT
The words, one row each (CoNLL-U)
#wordlemmaUPOSfeaturesheadrelation
1TwotwoNUMNumForm=Word NumType=Card2 botsnummod
2botsbotNOUNNumber=Plur3 disagreensubj
3disagreedisagreeVERBMood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Finrootroot
4aboutaboutADP_6 tallcase
5howhowADVPronType=Int6 talladvmod
6talltallADJDegree=Pos3 disagreeobl
7aaDETDefinite=Ind PronType=Art8 mountaindet
8mountainmountainNOUNNumber=Sing6 tallnsubj
9isbeAUXMood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin6 tallcop
10..PUNCT_3 disagreepunct
#verbtense, moodwhowhat, whomto whomwherewhenotherroles by
1disagreepresentTwo botsvn correspond-36.1 «NP V»: agent=Agent
2ispresenta mountaintall propertytree rules

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 that tell you?

objauxnsubjiobjpunctrootWhatPRONdoesAUXthatPRONtellVERByouPRON?PUNCT
The words, one row each (CoNLL-U)
#wordlemmaUPOSfeaturesheadrelation
1WhatwhatPRONPronType=Int4 tellobj
2doesdoAUXMood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin4 tellaux
3thatthatPRONNumber=Sing PronType=Dem4 tellnsubj
4telltellVERBVerbForm=Infrootroot
5youyouPRONCase=Acc Person=2 PronType=Prs4 telliobj
6??PUNCT_4 tellpunct
#verbtense, moodwhowhat, whomto whomwherewhenotherroles by
1tell speech: assertpresent · do-supportthatyouvn transfer_mesg-37.1.1-1-1 «NP V NP-Dative NP»: agent=Agent you=Recipient what=Topic

The options

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

It's a sign to check a trusted source

nsubjcopdetmarkacldetamodobjrootItPRON'sAUXaDETsignNOUNtoPARTcheckVERBaDETtrustedVERBsourceNOUN
The words, one row each (CoNLL-U)
#wordlemmaUPOSfeaturesheadrelation
1ItitPRONCase=Nom Gender=Neut Number=Sing Person=3 PronType=Prs4 signnsubj
2'sbeAUXMood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin4 signcop
3aaDETDefinite=Ind PronType=Art4 signdet
4signsignNOUNNumber=Singrootroot
5totoPART_6 checkmark
6checkcheckVERBVerbForm=Inf4 signacl
7aaDETDefinite=Ind PronType=Art9 sourcedet
8trustedtrustVERBTense=Past VerbForm=Part Voice=Pass9 sourceamod
9sourcesourceNOUNNumber=Sing6 checkobj
#verbtense, moodwhowhat, whomto whomwherewhenotherroles by
1is apresentIta signtree rules
2checkto-infinitivea trusted sourcevn assessment-34.1 «NP V NP»: agent=Agent source=Theme
3trustparticiplevn conjecture-29.5-1 «NP V that S»: agent=Agent

Option 2

One of them must be broken and needs fixing

nsubj:passcasenmodauxaux:passccconjobjrootOneNUMofADPthemPRONmustAUXbeAUXbrokenVERBandCCONJneedsVERBfixingNOUN
The words, one row each (CoNLL-U)
#wordlemmaUPOSfeaturesheadrelation
1OneoneNUMNumForm=Word NumType=Card6 brokennsubj:pass
2ofofADP_3 themcase
3themtheyPRONCase=Acc Number=Plur Person=3 PronType=Prs1 Onenmod
4mustmustAUXVerbForm=Fin6 brokenaux
5bebeAUXVerbForm=Inf6 brokenaux:pass
6brokenbreakVERBTense=Past VerbForm=Part Voice=Passrootroot
7andandCCONJ_8 needscc
8needsneedVERBMood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin6 brokenconj
9fixingfixingNOUNNumber=Sing8 needsobj
#verbtense, moodwhowhat, whomto whomwherewhenotherroles by
1breakmust · passiveOne of themvn hurt-40.8.3-1-1 «NP V NP»: subject=Patient
2needpresentfixingvn want-32.1-1-1 «NP V NP ADV»: agent=Pivot fixing=Theme

Option 3

The mountain keeps changing its height every year

detnsubjxcompnmod:possobjdetobl:unmarkedrootTheDETmountainNOUNkeepsVERBchangingVERBitsPRONheightNOUNeveryDETyearNOUN
The words, one row each (CoNLL-U)
#wordlemmaUPOSfeaturesheadrelation
1ThetheDETDefinite=Def PronType=Art2 mountaindet
2mountainmountainNOUNNumber=Sing3 keepsnsubj
3keepskeepVERBMood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Finrootroot
4changingchangeVERBVerbForm=Ger3 keepsxcomp
5itsitsPRONCase=Gen Gender=Neut Number=Sing Person=3 Poss=Yes PronType=Prs6 heightnmod:poss
6heightheightNOUNNumber=Sing4 changingobj
7everyeveryDETPronType=Tot8 yeardet
8yearyearNOUNNumber=Sing4 changingobl:unmarked
#verbtense, moodwhowhat, whomto whomwherewhenotherroles by
1keeppresentthe mountaintree rules
2changegerundthe mountain not said heremountain's heightyearvn dress-41.1.1-1-1 «NP V NP»: agent=Agent height=Patient

Option 4

Both are right in their own different way

nsubjcopcasenmod:possamodamodoblrootBothDETareAUXrightADJinADPtheirPRONownADJdifferentADJwayNOUN
The words, one row each (CoNLL-U)
#wordlemmaUPOSfeaturesheadrelation
1BothbothDETPronType=Tot3 rightnsubj
2arebeAUXMood=Ind Number=Plur Person=3 Tense=Pres VerbForm=Fin3 rightcop
3rightrightADJDegree=Posrootroot
4ininADP_8 waycase
5theirtheirPRONCase=Gen Number=Plur Person=3 Poss=Yes PronType=Prs8 waynmod:poss
6ownownADJDegree=Pos8 wayamod
7differentdifferentADJDegree=Pos8 wayamod
8waywayNOUNNumber=Sing3 rightobl
#verbtense, moodwhowhat, whomto whomwherewhenotherroles by
1ispresentBoththeir own different wayright propertytree rules

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/t5/