Optimal Query Allocation in Extractive QA with LLMs: A Learning-to-Defer Framework with Theoretical Guarantees

TL;DR

This paper introduces a Learning-to-Defer framework for extractive question answering with LLMs, optimizing query allocation to improve answer accuracy and reduce computational costs with theoretical guarantees.

Contribution

It presents a novel allocation strategy that balances performance and cost, supported by theoretical guarantees, and demonstrates effectiveness on multiple QA datasets.

Findings

Improves answer reliability in extractive QA tasks.

Reduces computational overhead significantly.

Provides theoretical guarantees on deferral strategy.

Abstract

Large Language Models excel in generative tasks but exhibit inefficiencies in structured text selection, particularly in extractive question answering. This challenge is magnified in resource-constrained environments, where deploying multiple specialized models for different tasks is impractical. We propose a Learning-to-Defer framework that allocates queries to specialized experts, ensuring high-confidence predictions while optimizing computational efficiency. Our approach integrates a principled allocation strategy with theoretical guarantees on optimal deferral that balances performance and cost. Empirical evaluations on SQuADv1, SQuADv2, and TriviaQA demonstrate that our method enhances answer reliability while significantly reducing computational overhead, making it well-suited for scalable and efficient EQA deployment.