Fine-Tuning Large Language Models Using Entity Hallucination Index for Text Summarization
January 9th, 2026
We propose a reinforcement-learning-based fine-tuning approach for large language models that uses the Enti Hallucination Index (EHI) as a reward signal to reduce entity-level hallucination in text summarization. Experiments on meeting transcripts show that this method improves entity faithfulness and accuracy.
This study uses the Entity Hallucination Index as a reward signal to reduce incorrect entity generation in AI-summarized text. Current metrics are two coarse, this protocol employs fine-grained entity rewards to improve factuality without human labels. To begin, use two primary corpora, including a dialogue transcription data set of multi-turn meeting transcripts and the XSum News summarization benchmark.
Clean each data set, flatten the data, and partition it into 80%training, 10%validation, and 10%test sets. Perform tokenization and entity extraction using a transformer-based NER model D slim/BERT-based NER to ensure robust grounding. Generate initial summaries in zero shot mode using beam search with a width of four to estimate baseline hallucination prevalence.
Compute the entity hallucination index as a reference-free entity level metric that categorizes entities into five factors to quantify faithfulness. Reward extractiveness and positive hallucinations and penalize negative hallucinations, over-focused entities, and lost focus. Define the reward function to be proportional to entity faithfulness and maximize the expected EHI reward across generated summaries.
Apply a reinforced style update mechanism to optimize model parameters toward maximizing entity faithfulness. Use low rank adaptation to enable parameter efficient fine tuning and update only the LoRA adapters while keeping the base model weights frozen. Then apply policy gradient updates to encourage higher reward outputs and regenerate summaries for every 500 updates to refresh reward estimates.
Finally, generate final summaries using the fine-tuned models and compare them against baseline results. Evaluate models for informativeness using Rouge one, Rouge two, and Rouge L, fluency and factual consistency using fact CC and CAGs. EHI fine tuning reduced hallucinations and improved factuality across models and data sets with rightward shifts in EHI, and fact CC distributions confirming increased entity faithfulness and factual consistency, while CAGS scores remained largely unchanged.
The harmful hallucination types of negative, over focus, and lost focus were reduced in Mistral after fine-tuning, while the extractiveness factor almost doubled. Correlation analysis showed EHI and CAGS had a weak correlation in Mistral and DistilBART, but Flan-T5 showed a stronger positive correlation. Researchers can measure entity level faithfulness and hallucination rates in summaries using the automated EHI reward signal.
The most critical challenge is NER accuracy, as extraction errors can propagate into incorrect reward calculations during training. Future studies can integrate co-reference resolution and informativeness rewards to better balance faithfulness with summary coverage.
View the full transcript and gain access to thousands of scientific videos
Overview
This article presents a novel reward-driven fine-tuning framework for large language models (LLMs) aimed at reducing entity-level hallucinations in abstractive summarization. By leveraging the Entity Hallucination Index (EHI) as a guiding metric, the approach enhances the factual accuracy of generated summaries without sacrificing informativeness or generalizability.
Key Study Components
Area of Science
- Natural Language Processing
- Machine Learning
- Text Summarization
Background
- LLMs have significantly improved abstractive summarization quality.
- Entity-level hallucination, where incorrect or fabricated entities appear in summaries, remains a major challenge.
- Existing models often struggle to balance informativeness and factual accuracy.
- Metrics to directly penalize hallucinated entities are underexplored in model training.
Purpose of Study
- To develop a fine-tuning framework that reduces entity hallucination in summarization models.
- To introduce and utilize the Entity Hallucination Index (EHI) as a reward signal.
- To demonstrate improved factuality and robustness in LLM-generated summaries.
Methods Used
- Initial summaries generated using pre-trained models (e.g., Flan-T5, DistilBART, Mistral) on structured transcript datasets such as XSUM.
- Named entity extraction from both generated summaries and gold reference summaries.
- Calculation of EHI by evaluating entity precision and penalizing fabricated entities.
- Reinforcement learning fine-tuning using a REINFORCE-style update mechanism with EHI as the reward signal.
Main Results
- Fine-tuned models exhibited lower rates of entity hallucination.
- Informativeness of summaries was maintained despite reduced hallucination.
- EHI-guided models demonstrated better generalization on out-of-domain summarization tasks.
- Enhanced robustness and factual accuracy were observed in experimental evaluations.
Conclusions
- The EHI-based reward framework effectively reduces entity-level hallucinations in LLM summarization.
- Factuality and informativeness can be jointly optimized using this approach.
- The method offers a practical direction for improving the reliability of automated summarization systems.
Frequently Asked Questions
Executive Industry Relevance
Entity-level hallucination in large language model (LLM) summarization poses a significant risk to data integrity and decision-making in biopharma R&D. The Entity Hallucination Index (EHI)-guided fine-tuning framework directly addresses this challenge, enabling more reliable extraction of factual insights from scientific and operational text sources. Enhanced factuality in automated summarization supports higher predictive confidence and reduces downstream risk in knowledge-driven pipelines.
Strategic Applications in Biopharma R&D
Early Discovery & Target Validation
- Improves accuracy of literature and data mining for target identification by minimizing fabricated entities.
- Supports mechanistic de-risking by ensuring extracted entities reflect true biological relationships.
- Enables more confident hypothesis generation from automated text summarization outputs.
Screening & Assay Development
- Facilitates standardized extraction of assay targets and conditions from diverse documentation sources.
- Reduces risk of propagating incorrect entity information into screening workflows.
- Supports reproducibility by maintaining entity fidelity across automated summaries.
Translational & Preclinical Research
- Enhances mapping of preclinical findings to disease-relevant entities by reducing hallucination.
- Improves continuity of entity tracking from discovery through preclinical validation phases.
- Supports risk-adjusted advancement by providing more reliable entity-level evidence.
Pipeline & Workflow Integration
The EHI-guided fine-tuning framework integrates into the informatics layer of discovery, screening, and translational research pipelines, supporting robust entity extraction and summarization.
- Discovery Biology: Enables hypothesis testing and pathway clarification by ensuring entity accuracy in mined summaries.
- Screening: Provides reproducible, entity-faithful summaries for assay documentation and compound evaluation.
- Analytics: Delivers quantitative EHI scores to benchmark and compare summarization outputs across conditions.
- Translational Research: Maintains entity fidelity for biomarker and disease mapping when moving from discovery to preclinical stages.
- Enterprise Reuse: Establishes a scalable, reward-driven summarization capability for diverse biopharma text sources.
Operational & Enterprise Impact
- Scientific Value: Increases predictive confidence and reduces mechanistic ambiguity in text-derived insights.
- Operational Value: Standardizes entity extraction and summary generation across teams and projects.
- Strategic Value: Supports better go/no-go decisions by improving the reliability of automated knowledge extraction.
- Portfolio Impact: Enables risk-adjusted prioritization by reducing the propagation of incorrect entity data.
Implementation Considerations
- Requires expertise in natural language processing and reinforcement learning for model fine-tuning.
- Needs access to structured transcript datasets and gold-standard reference summaries for EHI computation.
- Demands cross-team agreement on entity definitions and extraction standards.
- May require adaptation of EHI thresholds for different biomedical domains or document types.
- Dependent on the quality of initial pre-trained models and reference data for optimal performance.
