Beyond Liars' Bench: The Impact of Lie Typology, Depth, and Sparsity on Deception Detection in LLMs

• URL: http://arxiv.org/abs/2607.20479v1
• Date: Fri, 29 May 2026 12:44:01 GMT
• Title: Beyond Liars' Bench: The Impact of Lie Typology, Depth, and Sparsity on Deception Detection in LLMs
• Abstract summary: Training probes to detect deceptive outputs from large language models is still an open problem.<n>Recent work has demonstrated that detection probes fail especially in out-of-domain scenarios.
• Score: 2.644306510353322
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Training probes to detect deceptive outputs from large language models is still an open problem. Recent work has demonstrated that detection probes fail especially in out-of-domain scenarios -- training on one type of lie does not transfer well to deception scenarios involving other types of lies. In this work, we conduct a systematic study on how various factors impact detection performance: representation depth, probe expressivity, sparse feature representations, and the lie typology of the training data. To this end, we augment standard benchmark training data with a supplementary dataset containing diverse types of deception, including fabrication, omission, and exaggeration examples. Analyzing these factors across seven probe types, our experimental results show that the optimal representation depth is highly dataset-dependent, more expressive probes provide only selective gains over linear baselines, and sparse autoencoder features perform similarly to dense hidden states. Ultimately, we demonstrate that the choice of training data and lie typology substantially changes detectability, highlighting that deception detection is a highly representation-dependent problem.

Related papers

• Foundation Models are Implicit Deepfake Detectors [8.86239184578671]
  In this work, we uncover a surprisingly consistent phenomenon: across multiple pretrained models, datasets, and both image and video domains, fake samples systematically produce lower-magnitude representations than their real counterparts.<n>Motivated by this finding, we formulate deepfake detection as an anomaly detection problem and show that simple statistics of feature magnitude achieve competitive performance with far more sophisticated deepfake detection methods.
  arXiv  Detail & Related papers  (2026-08-10T10:55:27Z)
• Building Better Deception Probes Using Targeted Instruction Pairs [1.610762469264735]
  Linear probes are a promising approach for monitoring AI systems for deceptive behaviour.<n>In this paper, we identify the importance of the instruction pair used during training.<n>We show that targeting specific deceptive behaviors through a human-interpretable taxonomy of deception leads to improved results on evaluation datasets.
  arXiv  Detail & Related papers  (2026-02-01T20:18:11Z)
• When Personalization Tricks Detectors: The Feature-Inversion Trap in Machine-Generated Text Detection [64.23509202768945]
  We introduce dataset, the first benchmark for evaluating detector robustness in personalized settings.<n>Our experimental results demonstrate large performance gaps across detectors in personalized settings.<n>We propose method, a simple and reliable way to predict detector performance changes in personalized settings.
  arXiv  Detail & Related papers  (2025-10-14T13:10:23Z)
• Spatial-Temporal-Spectral Unified Modeling for Remote Sensing Dense Prediction [20.1863553357121]
  Current deep learning architectures for remote sensing are fundamentally rigid.<n>We introduce the Spatial-Temporal-Spectral Unified Network (STSUN) for unified modeling.<n> STSUN can adapt to input and output data with arbitrary spatial sizes, temporal lengths, and spectral bands.<n>It unifies various dense prediction tasks and diverse semantic class predictions.
  arXiv  Detail & Related papers  (2025-05-18T07:39:17Z)
• Downstream-Pretext Domain Knowledge Traceback for Active Learning [138.02530777915362]
  We propose a downstream-pretext domain knowledge traceback (DOKT) method that traces the data interactions of downstream knowledge and pre-training guidance. DOKT consists of a traceback diversity indicator and a domain-based uncertainty estimator. Experiments conducted on ten datasets show that our model outperforms other state-of-the-art methods.
  arXiv  Detail & Related papers  (2024-07-20T01:34:13Z)
• Transcending Forgery Specificity with Latent Space Augmentation for Generalizable Deepfake Detection [57.646582245834324]
  We propose a simple yet effective deepfake detector called LSDA. It is based on a idea: representations with a wider variety of forgeries should be able to learn a more generalizable decision boundary. We show that our proposed method is surprisingly effective and transcends state-of-the-art detectors across several widely used benchmarks.
  arXiv  Detail & Related papers  (2023-11-19T09:41:10Z)
• Automated Deception Detection from Videos: Using End-to-End Learning Based High-Level Features and Classification Approaches [0.0]
  We propose a multimodal approach combining deep learning and discriminative models for deception detection. We employ convolutional end-to-end learning to analyze gaze, head pose, and facial expressions. Our approach is evaluated on five datasets, including a new Rolling-Dice Experiment motivated by economic factors.
  arXiv  Detail & Related papers  (2023-07-13T08:45:15Z)
• On the Universal Adversarial Perturbations for Efficient Data-free Adversarial Detection [55.73320979733527]
  We propose a data-agnostic adversarial detection framework, which induces different responses between normal and adversarial samples to UAPs. Experimental results show that our method achieves competitive detection performance on various text classification tasks.
  arXiv  Detail & Related papers  (2023-06-27T02:54:07Z)
• Reinforcement Guided Multi-Task Learning Framework for Low-Resource Stereotype Detection [3.7223111129285096]
  "Stereotype Detection" datasets mainly adopt a diagnostic approach toward large Pre-trained Language Models. Annotating a reliable dataset requires a precise understanding of the subtle nuances of how stereotypes manifest in text. We present a multi-task model that leverages the abundance of data-rich neighboring tasks to improve the empirical performance on "Stereotype Detection"
  arXiv  Detail & Related papers  (2022-03-27T17:16:11Z)
• Learning What Makes a Difference from Counterfactual Examples and Gradient Supervision [57.14468881854616]
  We propose an auxiliary training objective that improves the generalization capabilities of neural networks. We use pairs of minimally-different examples with different labels, a.k.a counterfactual or contrasting examples, which provide a signal indicative of the underlying causal structure of the task. Models trained with this technique demonstrate improved performance on out-of-distribution test sets.
  arXiv  Detail & Related papers  (2020-04-20T02:47:49Z)
• Utilizing Network Properties to Detect Erroneous Inputs [0.76146285961466]
  We train a linear SVM classifier to detect erroneous data using hidden and softmax feature vectors of pre-trained neural networks. Our results indicate that these faulty data types generally exhibit linearly separable activation properties from correct examples. We experimentally validate our findings across a diverse range of datasets, domains, pre-trained models, and adversarial attacks.
  arXiv  Detail & Related papers  (2020-02-28T03:20:55Z)
• Stance Detection Benchmark: How Robust Is Your Stance Detection? [65.91772010586605]
  Stance Detection (StD) aims to detect an author's stance towards a certain topic or claim. We introduce a StD benchmark that learns from ten StD datasets of various domains in a multi-dataset learning setting. Within this benchmark setup, we are able to present new state-of-the-art results on five of the datasets.
  arXiv  Detail & Related papers  (2020-01-06T13:37:51Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.