# TPA: Next Token Probability Attribution for Detecting Hallucinations in RAG

arXiv: https://arxiv.org/abs/2512.07515 · HTML: https://mlsanity.com/paper/2512.07515/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Hallucination detection in RAG, one of 15 directions in Retrieval-augmented generation.
- Builds on 20 of its 53 references that are in the map.
- First posted on arXiv 8 Dec 2025.

## Placed in

- Retrieval-augmented generation / [Hallucination detection in RAG](https://mlsanity.com/rag/hallucination-detection-in-rag/)

## Builds on (most cited first, 20 of 20)

- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [XGBoost: A Scalable Tree Boosting System](https://mlsanity.com/paper/1603.02754/) (2016)
- [Language Models (Mostly) Know What They Know](https://mlsanity.com/paper/2207.05221/) (2022)
- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [Energy-based Out-of-distribution Detection](https://mlsanity.com/paper/2010.03759/) (2020)
- [Discovering Latent Knowledge in Language Models Without Supervision](https://mlsanity.com/paper/2212.03827/) (2022)
- [RAGTruth: A Hallucination Corpus for Developing Trustworthy Retrieval-Augmented Language Models](https://mlsanity.com/paper/2401.00396/) (2023)
- [Semantic Entropy Probes: Robust and Cheap Hallucination Detection in LLMs](https://mlsanity.com/paper/2406.15927/) (2024)
- [Out-of-Distribution Detection and Selective Generation for Conditional Language Models](https://mlsanity.com/paper/2209.15558/) (2022)
- [LM vs LM: Detecting Factual Errors via Cross Examination](https://mlsanity.com/paper/2305.13281/) (2023)
- [ReDeEP: Detecting Hallucination in Retrieval-Augmented Generation via Mechanistic Interpretability](https://mlsanity.com/paper/2410.11414/) (2024)
- [Enhancing Uncertainty-Based Hallucination Detection with Stronger Focus](https://mlsanity.com/paper/2311.13230/) (2023)
- [Steer LLM Latents for Hallucination Detection](https://mlsanity.com/paper/2503.01917/) (2025)
- [Chainpoll: A high efficacy method for LLM hallucination detection](https://mlsanity.com/paper/2310.18344/) (2023)
- [InterrogateLLM: Zero-Resource Hallucination Detection in LLM-Generated Answers](https://mlsanity.com/paper/2403.02889/) (2024)
- [Trust Me, I'm Wrong: LLMs Hallucinate with Certainty Despite Knowing the Answer](https://mlsanity.com/paper/2502.12964/) (2025)
- [Online Boosting Adaptive Learning under Concept Drift for Multistream Classification](https://mlsanity.com/paper/2312.10841/) (2023)
- [Drift-aware Collaborative Assistance Mixture of Experts for Heterogeneous Multistream Learning](https://mlsanity.com/paper/2508.01598/) (2025)
- [From Newborn to Impact: Bias-Aware Citation Prediction](https://mlsanity.com/paper/2510.19246/) (2025)
- [Learning to Reason for Hallucination Span Detection](https://mlsanity.com/paper/2510.02173/) (2025)

## Built on by (most cited first, 2 of 2)

- [Memory in the Age of AI Agents](https://mlsanity.com/paper/2512.13564/) (2025)
- [From Query to Counsel: Structured Reasoning with a Multi-Agent Framework and Dataset for Legal Consultation](https://mlsanity.com/paper/2604.10470/) (2026)

## Abstract (arXiv)

Detecting hallucinations in Retrieval-Augmented Generation remains a challenge. Prior approaches attribute hallucinations to a binary conflict between internal knowledge stored in FFNs and the retrieved context. However, this perspective is incomplete, failing to account for the impact of other components of the LLM, such as the user query, previously generated tokens, the self token, and the final LayerNorm adjustment. To comprehensively capture the impact of these components on hallucination detection, we propose TPA which mathematically attributes each token's probability to seven distinct sources: Query, RAG Context, Past Token, Self Token, FFN, Final LayerNorm, and Initial Embedding. This attribution quantifies how each source contributes to the generation of the next token. Specifically, we aggregate these attribution scores by Part-of-Speech (POS) tags to quantify the contribution of each model component to the generation of specific linguistic categories within a response. By leveraging these patterns, such as detecting anomalies where Nouns rely heavily on LayerNorm, TPA effectively identifies hallucinated responses. Extensive experiments show that TPA achieves state-of-the-art performance.