# HALT-RAG: A Task-Adaptable Framework for Hallucination Detection with Calibrated NLI Ensembles and Abstention

arXiv: https://arxiv.org/abs/2509.07475 · HTML: https://mlsanity.com/paper/2509.07475/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Builds on 7 of its 14 references that are in the map.
- First posted on arXiv 9 Sep 2025.

## Builds on (most cited first, 7 of 7)

- [Evaluating the Factual Consistency of Abstractive Text Summarization](https://mlsanity.com/paper/1910.12840/) (2019)
- [SummaC: Re-Visiting NLI-based Models for Inconsistency Detection in Summarization](https://mlsanity.com/paper/2111.09525/) (2021)
- [Asking and Answering Questions to Evaluate the Factual Consistency of Summaries](https://mlsanity.com/paper/2004.04228/) (2020)
- [FEQA: A Question Answering Evaluation Framework for Faithfulness Assessment in Abstractive Summarization](https://mlsanity.com/paper/2005.03754/) (2020)
- [Selective Question Answering under Domain Shift](https://mlsanity.com/paper/2006.09462/) (2020)
- [Q2: Evaluating Factual Consistency in Knowledge-Grounded Dialogues via Question Generation and Question Answering](https://mlsanity.com/paper/2104.08202/) (2021)
- [FaithDial: A Faithful Benchmark for Information-Seeking Dialogue](https://mlsanity.com/paper/2204.10757/) (2022)

## Built on by (most cited first, 3 of 3)

- [PassiveQA: A Three-Action Framework for Epistemically Calibrated Question Answering via Supervised Finetuning](https://mlsanity.com/paper/2604.04565/) (2026)
- [Twin Worlds: Equivariance-Based Abstention for Evidence-Grounded Reasoning](https://mlsanity.com/paper/2608.28018/) (2026)
- [The Semantic Illusion: Certified Limits of Embedding-Based Hallucination Detection in RAG Systems](https://mlsanity.com/paper/2512.15068/) (2025)

## Abstract (arXiv)

Detecting content that contradicts or is unsupported by a given source text is a critical challenge for the safe deployment of generative language models. We introduce HALT-RAG, a post-hoc verification system designed to identify hallucinations in the outputs of Retrieval-Augmented Generation (RAG) pipelines. Our flexible and task-adaptable framework uses a universal feature set derived from an ensemble of two frozen, off-the-shelf Natural Language Inference (NLI) models and lightweight lexical signals. These features are used to train a simple, calibrated, and task-adapted meta-classifier. Using a rigorous 5-fold out-of-fold (OOF) training protocol to prevent data leakage and produce unbiased estimates, we evaluate our system on the HaluEval benchmark. By pairing our universal feature set with a lightweight, task-adapted classifier and a precision-constrained decision policy, HALT-RAG achieves strong OOF F1-scores of 0.7756, 0.9786, and 0.7391 on the summarization, QA, and dialogue tasks, respectively. The system's well-calibrated probabilities enable a practical abstention mechanism, providing a reliable tool for balancing model performance with safety requirements.