Faithfulness-QA: A Counterfactual Entity Substitution Dataset for Training Context-Faithful RAG Models

2026-04-28

AI Summary

This work addresses the tendency of retrieval-augmented generation (RAG) models to disregard retrieved context in favor of internal parametric knowledge, leading to unfaithful outputs. To mitigate this issue, the authors propose a type-consistent counterfactual entity replacement method to construct a large-scale, controllable knowledge-conflict dataset comprising 99,094 samples. This dataset compels models to prioritize contextual information over internal knowledge when conflicts arise. The construction pipeline integrates automatic named entity recognition, type-aligned entity substitution, rigorous quality filtering, and an attention-guided training objective designed to enhance contextual faithfulness. In a random audit of 200 samples, the dataset achieved a 100% pass rate across four automated consistency checks, demonstrating its effectiveness in improving RAG models’ adherence to provided context and supporting both training and evaluation of faithful generation systems.

Abstract

Retrieval-Augmented Generation (RAG) models frequently produce answers grounded in parametric memory rather than the retrieved context, undermining the core promise of retrieval augmentation. A fundamental obstacle to fixing this unfaithfulness is the lack of training data that explicitly requires models to prefer context over internal knowledge. We introduce Faithfulness-QA, a large-scale dataset of 99,094 samples constructed through counterfactual entity substitution. Starting from two established extractive QA benchmarks--SQuAD and TriviaQA--we automatically identify answer-bearing named entities in each context, replace them with type-consistent alternatives drawn from a curated bank of 76,953 entities, and thereby manufacture controlled knowledge conflicts between context and parametric memory. Rigorous quality filtering ensures 100% pass rates across four automated checks on random 200-sample audits. We release the full dataset, the construction pipeline, and a typed entity bank covering eight named entity categories. Faithfulness-QA is designed as a training resource for attention-based faithfulness objectives and as an evaluation benchmark for measuring context-grounding behavior in RAG systems. Data and code are available at https://github.com/qzhangFDU/faithfulness-qa-dataset.

Problem

• Retrieval-Augmented Generation
• faithfulness
• context grounding
• counterfactual entity substitution
• training data

Innovation

• Counterfactual Entity Substitution
• Retrieval-Augmented Generation
• Context Faithfulness
• Named Entity Replacement
• Faithfulness-QA