FaithLens: Detecting and Explaining Faithfulness Hallucination

TL;DR

FaithLens introduces a cost-efficient, explainable approach to detecting faithfulness hallucinations in LLM outputs by jointly predicting fidelity and generating explanations. It combines a cold-start supervised fine-tuning stage with a data-filtered synthetic dataset and a subsequent rule-based reinforcement learning stage to optimize both prediction accuracy and explanation quality. Across 12 tasks, FaithLens achieves state-of-the-art results with an 8B-parameter model and substantially lower cost than API-based LLMs, while delivering high-quality, actionable explanations. The work highlights the practicality of explainable detection for real-world LLM deployments and outlines limitations and future extensions, including multi-modal grounding and finer-grained hallucination taxonomy.

Abstract

Recognizing whether outputs from large language models (LLMs) contain faithfulness hallucination is crucial for real-world applications, e.g., retrieval-augmented generation and summarization. In this paper, we introduce FaithLens, a cost-efficient and effective faithfulness hallucination detection model that can jointly provide binary predictions and corresponding explanations to improve trustworthiness. To achieve this, we first synthesize training data with explanations via advanced LLMs and apply a well-defined data filtering strategy to ensure label correctness, explanation quality, and data diversity. Subsequently, we fine-tune the model on these well-curated training data as a cold start and further optimize it with rule-based reinforcement learning, using rewards for both prediction correctness and explanation quality. Results on 12 diverse tasks show that the 8B-parameter FaithLens outperforms advanced models such as GPT-4.1 and o3. Also, FaithLens can produce high-quality explanations, delivering a distinctive balance of trustworthiness, efficiency, and effectiveness.

Figures (17)

• Figure 1: The illustration of our FaithLens. Given a document $doc$ and a claim $c$, FaithLens can jointly determine whether the claim is faithful or hallucinated and provide the corresponding explanations for its decisions, applicable across various tasks.
• Figure 2: The Overall Process of Training FaithLens, including (1) Cold-Start SFT: We first synthesize high-quality data with explanations used for the SFT stage. (2) Rule-Based RL Training: We further refine the model using a rule-based RL approach with the designed rewards for both prediction correctness and explanation quality.
• Figure 3: Human Evaluation. We compare the explanations from FaithLens and GPT-4o on 120 samples.
• Figure 4: Prompt used for training and inference of FaithLens.
• Figure 5: Prompt used for data synthesis.