A language-model-only, two-phase retrieval pipeline with hard-negative training and a grid-searched ensemble is reported to outperform sparse, dense, and generative baselines on a Japanese legal retrieval test set and a small MS MARCO subset.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2024 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Optimizing Multi-Stage Language Models for Effective Text Retrieval
A language-model-only, two-phase retrieval pipeline with hard-negative training and a grid-searched ensemble is reported to outperform sparse, dense, and generative baselines on a Japanese legal retrieval test set and a small MS MARCO subset.