Reducing Hallucinations in Summarization via Reinforcement Learning with Entity Hallucination Index

TL;DR

This paper presents a reinforcement learning-based fine-tuning method that improves the factual accuracy of summaries by reducing entity hallucinations, using an automatic entity grounding metric called EHI, without human annotations.

Contribution

It introduces a scalable, reward-driven fine-tuning framework that explicitly optimizes for entity grounding in summarization models, reducing hallucinations effectively.

Findings

Significant reduction in entity hallucinations across datasets.

Improved entity correctness without sacrificing fluency.

Reproducible pipeline released for further research.

Abstract

Reducing hallucinations in abstractive summarization remains a critical challenge for deploying language models (LMs) in real-world settings. In this work, we introduce a rewarddriven fine-tuning framework that explicitly optimizes for Entity Hallucination Index (EHI), a metric designed to quantify the presence, correctness, and grounding of named entities in generated summaries. Given a corpus of meeting transcripts, we first generate baseline summaries using a pre-trained LM and compute EHI scores via automatic entity extraction and matching. We then apply reinforcement learning to fine-tune the model parameters, using EHI as a reward signal to bias generation toward entity-faithful outputs. Our approach does not rely on human-written factuality annotations, enabling scalable fine-tuning. Experiments demonstrate consistent improvements in EHI across datasets, with qualitative analysis revealing a significant reduction in entity-level hallucinations without degradation in fluency or informativeness. We release a reproducible Colab pipeline, facilitating further research on hallucination-aware model fine-tuning using lightweight, hallucintion metrics like EHI.