Guide-yakabatana mibvunzo · Zvakaoma Nhanho

ColBERT uye Multi-Vector Retrieval Quiz

Dzidza ColBERT's akawanda-vector chinzvimbo, kukwikwidzwa kwemumiriri, ColPali gwaro-peji kudzoreredza, uye kuti ColBERTv2 inomanikidza sei zviratidzo zvinomiririra.

Nzira dzekutungamira dzinoenderanaColbert uye Multi Vector Retrieval
Mubvunzo 1 ye 8

ColBERT inomiririra gwaro sei?

Tarisa mubvunzo wese8 mibvunzo

Mhinduro uye tsananguro zvinoonekwa paunenge uchitora bvunzo.

  1. ColBERT inomiririra gwaro sei?
    • Sezvo akawanda per-token contextual vectors
    • Seimwe dense vector
    • Sebhegi shoma remazwi akakosha
    • Setambo yemavara mbishi chete
  2. Ndeipi mashandiro anoshandiswa naColBERT kuwana zvine chekuita nemubvunzo-gwaro?
    • Cosine yemavheji maviri ega
    • MaxSim: hwerengedzo yemubvunzo wega wega tokeni inofananidzira kune chero gwaro tokeni
    • Yambuka-entropy pamusoro peshoko rose
    • BM25 term frequency
  3. Nei ColBERT ichikurumidza kupfuura muchinjiko-encoder pachiyero?
    • Inoshandisa mazwi mashoma
    • Iyo inofuratira mubvunzo zvachose
    • Document token embeddings precomputed uye indexed offline
    • Inogadzirisa chete izwi rekutanga regwaro rega rega
  4. Idambudziko ripi rine single-vector dense retriever rinoitwa naColBERT?
    • Vanononoka kunongedzera
    • Ivo havagone kushandisa neural network
    • Ivo havadi data yekudzidziswa
    • Ivo vanorasikirwa neyakaomeswa-yakaomeswa, tokeni-chikamu danho nekumanikidza kune imwe vector
  5. Ndeipi budiriro yakakosha yakaunzwa neColBERTv2?
    • Residual compression iyo inoderedza zvakanyanya saizi ye index
    • Yakabvisa BERT zvachose
    • Yakashandura kune imwe-vector encoding
    • Inoshanda chete pamifananidzo
  6. Chii chinonzi ColPali inoshandisa ColBERT-maitiro ekunonoka kusangana kune?
    • Audio waveforms
    • Zvimedu zvemifananidzo emapeji egwaro, senge maPDF
    • Single keywords chete
    • Database table rows
  7. In a two-stage search system, what role can ColBERT play after a fast first-stage retriever?
    • Generate the source documents from scratch
    • Replace the query with a document title
    • Rerank the candidate passages using contextual token-level matching
    • Discard every candidate before scoring
  8. What does ColBERTv2's residual compression store for a token vector?
    • Only the raw document text with no vector data
    • A full copy of the language model for each token
    • One shared vector for the entire document
    • A centroid code together with a compact residual representation