Researchllminformation retrievalbm25program evolution
RankEvolve Discovers Improved Lexical Retrieval Algorithms
8.1
Researchers introduce RankEvolve on Feb 18, 2026, an LLM-guided program-evolution framework that mutates, recombines, and selects executable ranking programs to improve lexical retrieval. Starting from BM25 and query likelihood seeds, the method is evaluated across 12 BEIR and BRIGHT datasets and transfers to full BEIR, BRIGHT, and TREC DL 19 and 20, demonstrating improved retrieval performance.
Key Points
- 1Evolves ranking programs using LLM-guided mutations and selection across 12 BEIR and BRIGHT datasets
- 2Demonstrates improved retrieval over BM25 and query likelihood seeds, indicating substantive algorithmic gains
- 3Enables practitioners to automatically discover and transfer novel lexical rankers to BEIR and TREC benchmarks
Scoring Rationale
Strong novelty and cross-benchmark transfer, but score limited by single-source arXiv preprint without peer review.
Sources
Public references used for this report.
Practice with real Logistics & Shipping data
90 SQL & Python problems · 15 industry datasets
Used by DS/ML engineers at top companies
High-Value Overnight OrdersEasyDelivered International ShipmentsMediumOn-Time Delivery Rate by CarrierHard
250 free problems · No credit card
See all Logistics & Shipping problems
