| name | 0.2.3-understand-rrf_rank_position |
| description | [0.2.3] The integer position of an item within a rank list, starting at 1 for the highest-scoring result; used as the |
understand-rrf_rank_position
CALL NUMBER: deep_dense_retrieval.rrf_rank_position
DEFINITION: The integer position of an item within a rank list, starting at 1 for the highest-scoring result; used as the denominator in reciprocal rank computation.
Invoke this skill to understand rrf_rank_position down to its primitives. The RELATIVE ROOT below is the least-fixed-point closure of everything it bundles from — the full import cone, grouped by the lib each prim comes from. Projected from a prover-typed KB (MAP/SWI-Prolog consistency gate): every reference below resolves.
THE RELATIVE ROOT (the import cone, by lib)
from deep_dense_retrieval
- rrf_constant_k (d1): A positive smoothing constant (conventional value 60) added to each rank position before taking the reciprocal, preventing division by zero and dampening sensitivity to high ranks for stable fusion scoring
- rrf_reciprocal_rank_score (d1): The per-system score for a document computed as 1 divided by the sum of rank position and constant k, yielding a value in (0, 1/k] that decays with worsening rank
- rrf_fusion_score (d2): The sum of reciprocal rank values for a given document across all contributing rank lists; the document's position in the fused ranking is determined by descending fusion score.
- rrf_final_ranking (d3): The merged document list produced by sorting all candidates in descending order by their fusion scores, yielding a unified cross-system ranking that reflects all input retrieval evidence
CONSUMERS (what needs this)
rrf_normalization_preprocessing, rrf_ranked_list
Projected from the retrieval augmented generation architecture patterns KB (207 concepts / 225 relations) — consistency-typed by MAP; the facet list after the colon IS the cross-lib dependency web.
(leaf — this is an actual skill.)
