Maswali ya Utafutaji Mseto

Majibu na maelezo huonekana unapojibu maswali.

1. Je, ni mbinu gani mbili za urejeshaji ambazo utafutaji wa mseto kawaida huchanganya?
   • Utafutaji wa maneno muhimu machache (k.m., BM25) na utafutaji mnene wa vekta
   • Aina mbili za lugha kubwa zinazopigia kura majibu
   • Hoja za SQL na faharasa za maandishi kamili pekee
   • Utafutaji wa picha na utaftaji wa sauti
2. Kwa nini Reciprocal Rank Fusion (RRF) hutumiwa kwa kawaida kuunganisha matokeo?
   • Hufunza mtandao mpya wa neva kwa kila hoja
   • Inachanganya hati kwa nafasi ya cheo badala ya alama ghafi, kuepuka kutofautiana kwa kiwango
   • Inafuta hati yoyote inayoonekana katika orodha zote mbili
   • Inaweka tu matokeo ya juu kutoka kwa kila mtoaji
3. Ni hali gani inayopendelea sehemu ndogo (neno kuu) ya utaftaji wa mseto?
   • Kutafuta hati zinazofafanua swali kwa maneno tofauti
   • Kuelewa sauti ya kihisia ya sentensi
   • Kulinganisha sehemu kamili ya nambari au msimbo wa hitilafu ambao lazima uonekane kihalisi
   • Kutafsiri swali katika lugha nyingine
4. Je, ni udhaifu gani mkuu wa kutumia utaftaji wa vekta mnene pekee?
   • Haiwezi kuelewa visawe au vifungu vya maneno
   • Inafanya kazi kwenye picha tu
   • Haihitaji upachikaji hata kidogo
   • Inaweza kukosa ishara nadra kabisa kama vile misimbo maalum au majina ya ukoo
5. Je! mfano wa sparse uliojifunza kama SPLADE hufanya nini?
   • Hutoa uzani wa maneno unaoweza kufasiriwa kwa kutumia mtandao wa neva, kuchanganya mawazo machache na mnene
   • Huondoa ulinganishaji wa maneno yote muhimu kutoka kwa utafutaji
   • Inahifadhi picha tu kama vekta
   • Inachukua nafasi ya haja ya kurejesha yoyote
6. What can combining lexical and dense retrieval provide to a RAG system?
   • A guarantee that every generated statement is true
   • Potentially more varied candidate evidence, subject to evaluation of retrieval and answer quality
   • A replacement for grounding sources
   • An automatic legal review of generated answers
7. What tradeoff did BEIR report for reranking and late-interaction methods on average?
   • They always ran faster and used less compute than BM25
   • They made retrieval independent of the corpus
   • They often achieved stronger zero-shot performance but at higher computational cost
   • They guaranteed a correct generated answer
8. What did the BEIR benchmark conclude about BM25 in zero-shot retrieval?
   • It always beats every neural retriever
   • It cannot retrieve exact terms
   • It is a dense-vector model
   • It remains a robust baseline across diverse datasets