RAG & Retrieval

Chunking, vector search, hybrid ranking, re-ranking and the checks that catch an ungrounded answer.

Intermediate · 21 problems · 0 passed · Background reading: Retrieval-Augmented Generation (RAG)

1. 01Fixed-Size Chunking With Overlapeasy
2. 02Cosine Similarity Between Vectorseasy
3. 03Sentence-Boundary-Aware Chunkingmedium
4. 04Brute-Force K-Nearest-Neighbor Searcheasy
5. 05Build an Inverted Indexeasy
6. 06Reciprocal Rank Fusionmedium
7. 07Dense + Sparse Score Fusionmedium
8. 08Context Window Token Budget Fittingmedium
9. 09Retrieval Recall@Kmedium
10. 10Scalar Quantize a Vectormedium
11. 11Binary Quantization and Hamming Distanceeasy
12. 12Merge Top-K Results From Multiple Shardsmedium
13. 13Filtered ANN Search With Metadata Predicatemedium
14. 14HNSW Greedy Search Single Hopmedium
15. 15Grounding Check via N-Gram Overlapmedium
16. 16Faithfulness Evaluation Scoremedium
17. 17Corrective RAG: Grade Retrieved Document Relevancemedium
18. 18Decompose a Compound Question Into Subquestionsmedium
19. 19BFS Shortest Path Between Two Entitiesmedium
20. 20Maximal Marginal Relevance (MMR) Selection Stephard
21. 21One PageRank Iteration Stephard

Next track: Agents in Production