# Training-free Truthfulness Detection via Sparse MLP Value Vectors

arXiv: https://arxiv.org/abs/2509.17932 · HTML: https://mlsanity.com/paper/2509.17932/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 26 of its 32 references that are in the map.
- First posted on arXiv 22 Sep 2025.

## Builds on (most cited first, 26 of 26)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [TruthfulQA: Measuring How Models Mimic Human Falsehoods](https://mlsanity.com/paper/2109.07958/) (2021)
- [PIQA: Reasoning about Physical Commonsense in Natural Language](https://mlsanity.com/paper/1911.11641/) (2019)
- [GLU Variants Improve Transformer](https://mlsanity.com/paper/2002.05202/) (2020)
- [Understanding intermediate layers using linear classifier probes](https://mlsanity.com/paper/1610.01644/) (2016)
- [Transformer Feed-Forward Layers Are Key-Value Memories](https://mlsanity.com/paper/2012.14913/) (2020)
- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [Discovering Latent Knowledge in Language Models Without Supervision](https://mlsanity.com/paper/2212.03827/) (2022)
- [The Internal State of an LLM Knows When It's Lying](https://mlsanity.com/paper/2304.13734/) (2023)
- [Knowledge Neurons in Pretrained Transformers](https://mlsanity.com/paper/2104.08696/) (2021)
- [Transformer Feed-Forward Layers Build Predictions by Promoting Concepts in the Vocabulary Space](https://mlsanity.com/paper/2203.14680/) (2022)
- [HaluEval: A Large-Scale Hallucination Evaluation Benchmark for Large Language Models](https://mlsanity.com/paper/2305.11747/) (2023)
- [SWAG: A Large-Scale Adversarial Dataset for Grounded Commonsense Inference](https://mlsanity.com/paper/1808.05326/) (2018)
- [QASC: A Dataset for Question Answering via Sentence Composition](https://mlsanity.com/paper/1910.11473/) (2019)
- [Adversarial GLUE: A Multi-Task Benchmark for Robustness Evaluation of Language Models](https://mlsanity.com/paper/2111.02840/) (2021)
- [Lookback Lens: Detecting and Mitigating Contextual Hallucinations in Large Language Models Using Only Attention Maps](https://mlsanity.com/paper/2407.07071/) (2024)
- [HaloScope: Harnessing Unlabeled LLM Generations for Hallucination Detection](https://mlsanity.com/paper/2409.17504/) (2024)
- [Attention Satisfies: A Constraint-Satisfaction Lens on Factual Errors of Language Models](https://mlsanity.com/paper/2309.15098/) (2023)
- [Do Androids Know They're Only Dreaming of Electric Sheep?](https://mlsanity.com/paper/2312.17249/) (2023)
- [Steer LLM Latents for Hallucination Detection](https://mlsanity.com/paper/2503.01917/) (2025)
- [CICERO: A Dataset for Contextualized Commonsense Inference in Dialogues](https://mlsanity.com/paper/2203.13926/) (2022)
- [On the Universal Truthfulness Hyperplane Inside LLMs](https://mlsanity.com/paper/2407.08582/) (2024)
- [Multiview Contextual Commonsense Inference: A New Dataset and Task](https://mlsanity.com/paper/2210.02890/) (2022)
- [Enhanced Hallucination Detection in Neural Machine Translation through Simple Detector Aggregation](https://mlsanity.com/paper/2402.13331/) (2024)

## Abstract (arXiv)

Large language models (LLMs) are prone to generating factually incorrect content, motivating methods for assessing truthfulness from internal model signals. While supervised probing approaches can be effective, they require labeled data and classifier training. Recent training-free methods avoid parameter optimization but rely on coarse activation statistics that provide limited insight into how truthfulness-related signals arise within the model. We present a training-free approach that operates at the level of individual multi-layer perceptron (MLP) value vectors. Through a systematic analysis, we find that although most value vectors show no meaningful signal, a sparse subset exhibits stable and directionally consistent correlations with content truthfulness. Leveraging this observation, we propose TruthV, a simple inference method that aggregates preferences expressed by these value vectors. TruthV requires only a small support set to identify relevant vectors and introduces no additional model parameters or classifier weights. We evaluate TruthV across model scales from 2B to 13B and multiple benchmarks, including question answering, natural language understanding, and hallucination evaluation. TruthV consistently outperforms existing training-free baselines, demonstrating that truthfulness-related variation in LLMs is captured in a sparse and structured manner at the level of MLP value vectors.