affb59fff6
RRF merge (k=60) benchmarked against ground truth with 10K queries on skewed 10-shard corpus (93% on shard 1). Result: Kendall τ = 0.1369 (95% CI [0.1339, 0.1399]), far below the 0.95 threshold. 9,998 of 10,000 queries fell below τ=0.95, confirming RRF alone is insufficient for cross-shard ranking quality with skewed distributions. DFS preflight (already implemented) achieves τ = 0.9818, passing the threshold. Add full 10K-query DFS comparison report and fix paths in experiment.json. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| corpus | ||
| queries | ||
| results | ||
| README.md | ||
| simulate.py | ||
Score Comparability Benchmark
Tests whether _rankingScore values from different shards are comparable when documents are distributed unevenly across shards.
Problem Statement
Meilisearch's ranking pipeline computes scores using local statistics (term frequency, document frequency). When shards have very different document distributions, identical queries may return scores that aren't directly comparable, leading to incorrect merged rankings.
Experiment Design
- Ground truth: Single Meilisearch index with all documents
- Distributed setup: Same documents sharded across N nodes with intentional skew
- Measurement: Kendall tau (τ) between merged distributed results and ground truth
- Pass criterion: τ ≥ 0.95 on average across 10k random queries
Corpus Structure
- 100,000 documents total
- 10 shards (shard 0 = normal, shard 1 = 100× normal, shard 9 = 0.01× normal)
- Documents have: id, title, content (synthetic text), category (for filtering)
- 50 unique terms distributed across documents with varying frequencies
Directory Layout
corpus/: Test document sets (JSONL)queries/: Generated query sets for experimentsresults/: Experimental results and analysis
Running Experiments
See individual experiment scripts in results/ directories.