Engineering / advanced
Rank evidence under a latency budget
Compare lexical, dense, hybrid, and learned ranking through explicit relevance judgments, candidate sets, and operating constraints.
Your route through engineering
Follow the path or begin with a question. Every lesson is open.
Your place, at your pace. Progress is optional and saved only in this browser when available.
Read / explore / reason
Inside this module
Module 01 advanced
Rank evidence under a latency budget
Compare lexical, dense, hybrid, and learned ranking through explicit relevance judgments, candidate sets, and operating constraints.
- 01
Lexical term weighting and BM25
Declare term-frequency saturation, field-length normalization, and IDF conventions before interpreting a lexical retrieval score.
27 min - 02
Query-document dual encoders
Align encoder versions, pooling, normalization, and search metric so an indexed vector means the same thing at training and inference.
27 min - 03
Hard-negative mining and false negatives
Mine informative confusions without equating unjudged passages with proven negatives or treating teacher labels as certainty.
27 min - 04
Cross-encoder reranking score semantics
Use joint query-document scoring within a candidate set while separating ranking quality, calibration, and candidate recall.
27 min - 05
Reciprocal-rank fusion and score combination
Combine ranked lists with explicit truncation and tie policies, or combine comparable features with validated scales and weights.
27 min - 06
Learning-to-rank objectives
Choose pointwise, pairwise, or listwise supervision with explicit query groups, label gains, ties, and evaluation alignment.
28 min - 07
Relevance judgments and pooling bias
Separate unjudged documents from nonrelevant ones and report how incomplete pools affect apparent ranking quality.
28 min - 08
Diversity and maximal marginal relevance
Greedily select useful nonredundant evidence while declaring relevance scales, similarity meaning, and the limits of MMR.
28 min - 09
Query routing and reranker budgets
Allocate retrieval branches and candidate depth within an explicit latency model while measuring routing errors and tail behavior.
28 min - 10
Ranking error analysis by query class
Localize ranking failures and uncertainty with per-query evidence, class denominators, paired comparisons, and stable judgment policies.
28 min