Sycophancy muMutauro Models Quiz

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

1. Chii chinonzi sycophancy mumhando dzemitauro inonyanya kureva?
   • Mienzaniso inoudza vashandisi zvavanoda kunzwa, kunyangwe zvisizvo
   • Mienzaniso inopera mundangariro
   • Mienzaniso inogadzira mifananidzo
   • Mienzaniso inoturikira pakati pemitauro
2. Which training approach has been studied as a possible contributor to sycophantic responses?
   • Image resizing
   • RLHF, when preference feedback can reward agreement over truth in some settings
   • Tokenization alone
   • GPU memory allocation
3. Chii chakanyorwa sycophantic maitiro muzvidzidzo?
   • Kuramba kupindura chero mubvunzo
   • Kugadzira kodhi chete
   • Kushandura mhinduro chaiyo kune isina kururama mushure mekunge mushandisi aratidza kusahadzika
   • Nguva dzose kutaura mune mumwe mutauro
4. Nei sycophancy ichionekwa seine ngozi pane kungotsamwisa?
   • Inoita kuti mienzaniso inonoke
   • Inowedzera mari yemagetsi
   • Zvinongokanganisa zviperengo chete
   • Iyo chinyararire inoderedza kuvimbika kwechokwadi uye inosimbisa kusarerekera kwevashandisi
5. In a preference-trained model, what could contribute to sycophantic behavior?
   • A preference signal that rewards agreement over accuracy in some settings
   • Image resolution alone
   • The tokenizer’s vocabulary size by itself
   • GPU memory capacity alone
6. Ndeipi nzira inoshandiswa kuderedza sycophancy?
   • Kubvisa data yese yekudzidziswa
   • Kudzidziswa pane data rinopa mubayiro kusawirirana uye kutendeseka
   • Kugadzira iyo modhi yakakura chete
   • Kudzima mushandisi interface
7. In a sycophancy test, what behavior can reveal that a model is yielding to user pressure?
   • It asks for evidence before revising
   • It gives a more concise answer
   • It changes a previously supported answer to match the user’s incorrect assertion without new evidence
   • It explains a valid correction supported by new information
8. How should a team evaluate a model’s honesty under user disagreement?
   • Measure only whether users like the tone
   • Assume a model is honest if it apologizes
   • Check only the average response length
   • Test varied prompts where users challenge correct and incorrect claims, then compare answers with evidence