M7
Hybrid Retrieval
Reciprocal rank fusion
A hybrid retriever that combines lexical and semantic signals
Available
Cohort
- Effort
- 3-5 hours
- Prerequisite
- M6
- Core concept
- Fusion and reranking
What you have
Two separate retrieval paths
What you gain
One ranked list combining both signals
What you build
The module is planned. It will combine the lexical and semantic retrieval paths into one ranked output.
- A fusion pipeline that combines lexical scores and vector scores
- Configurable weighting for retrieval signals
- Reranking logic for one final merged result list
What you learn
- How hybrid retrieval balances precision and recall
- Why signal fusion needs calibration rather than blind averaging
- How to inspect retrieval disagreements before merging scores
Artifact and workload
Primary artifact: Fusion pipeline with configurable weights
Tests20
Assessments4
Estimated time3-5 hours
Access
This module is part of the cohort. Join the guided path for reviews, deadlines, and the workshop sequence after the ranking modules.