Are Optimal Algorithms Still Optimal? Rethinking Sorting in LLM-Based Pairwise Ranking with Batching and Caching
About
We introduce a novel framework for analyzing sorting algorithms in pairwise ranking prompting (PRP), re-centering the cost model around LLM inferences rather than traditional pairwise comparisons. While classical metrics based on comparison counts have traditionally been used to gauge efficiency, our analysis reveals that expensive LLM inferences overturn these predictions; accordingly, our framework encourages strategies such as batching and caching to mitigate inference costs. We show that algorithms optimal in the classical setting can lose efficiency when LLM inferences dominate the cost under certain optimizations.
Juan Wisznia, Cecilia Bola\~nos, Juan Tollo, Giovanni Marraffini, Agust\'in Gianolini, Noe Hsueh, Luciano Del Corro• 2025
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Reranking | TREC DL 2020 | NDCG@100.7045 | 132 | |
| Reranking | TREC DL 2019 (test) | NDCG@1067.12 | 108 | |
| Document Reranking | TREC DL 2019 and 2020 (test) | NDCG@1065.42 | 108 | |
| Reranking | TREC DL 2019 v1 (test) | NDCG@1065.73 | 108 | |
| Document Reranking | TREC DL 19 | NDCG@1067.44 | 39 | |
| Reranking | BEIR (test) | Covid Score74.8 | 19 | |
| End-to-end Reranking | TREC DL 2019 2020 Average | Average NDCG@1068.95 | 10 |
Showing 7 of 7 rows