Prompts as Auto-Optimized Training Hyperparameters: Training Best-in-Class IR Models from Scratch with 10 Gold Labels

TL;DR

This paper introduces a method to train small neural IR models with minimal labeled data by automatically optimizing prompts for synthetic query generation, achieving competitive results with much larger models.

Contribution

The paper presents an automatic prompt optimization technique for generating synthetic queries, enabling effective training of small IR models with very limited labeled data.

Findings

Models trained with the method outperform RankZephyr.

The approach is competitive with RankLLama trained on more data.

Automatic prompt optimization enhances synthetic dataset quality.

Abstract

We develop a method for training small-scale (under 100M parameter) neural information retrieval models with as few as 10 gold relevance labels. The method depends on generating synthetic queries for documents using a language model (LM), and the key step is that we automatically optimize the LM prompt that is used to generate these queries based on training quality. In experiments with the BIRCO benchmark, we find that models trained with our method outperform RankZephyr and are competitive with RankLLama, both of which are 7B parameter models trained on over 100K labels. These findings point to the power of automatic prompt optimization for synthetic dataset generation.