Subliminal Corruption: Mechanisms, Thresholds, and Interpretability

• URL: http://arxiv.org/abs/2510.19152v1
• Date: Wed, 22 Oct 2025 00:57:51 GMT
• Title: Subliminal Corruption: Mechanisms, Thresholds, and Interpretability
• Abstract summary: Subliminal corruption is transmitted through semantically neutral data, bypassing standard safety checks.<n>We present a systematic study of the scaling laws, thresholds, and mechanisms of subliminal corruption using a teacher-student setup with GPT-2.<n>Results demonstrate a critical vulnerability in AI systems that rely on synthetic data and highlight the need for new safety protocols.
• Score: 0.6875312133832079
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: As machine learning models are increasingly fine-tuned on synthetic data, there is a critical risk of subtle misalignments spreading through interconnected AI systems. This paper investigates subliminal corruption, which we define as undesirable traits are transmitted through semantically neutral data, bypassing standard safety checks. While this phenomenon has been identified, a quantitative understanding of its dynamics is missing. To address this gap, we present a systematic study of the scaling laws, thresholds, and mechanisms of subliminal corruption using a teacher-student setup with GPT-2. Our experiments reveal three key findings: (1) subliminal corruption causes behavioral crossover, degrading the model's overall alignment, not just the targeted trait; (2) alignment fails in a sharp phase transition at a critical threshold of poisoned data, rather than degrading gradually; and (3) interpretability analysis shows the corruption mechanism mimics the model's natural fine-tuning process, making it difficult to detect. These results demonstrate a critical vulnerability in AI systems that rely on synthetic data and highlight the need for new safety protocols that can account for latent threats.

Related papers

• SafeCA: Safe Cross-Attention Localization and Regulation for Text-to-Video Jailbreak Defense [77.70598852234149]
  Text-to-Video (T2V) generative models are vulnerable to jailbreak attacks in real-world deployment.<n>We propose SafeCA, a feature-level defense mechanism for safe cross-attention localization and regularization.<n> Experimental results show that SafeCA reduces the jailbreak success rate by about 20% on mainstream T2V models.
  arXiv  Detail & Related papers  (2026-08-11T14:01:24Z)
• Character as a Latent Variable in Large Language Models: A Mechanistic Account of Emergent Misalignment and Conditional Safety Failures [70.48661957773449]
  Emergent Misalignment refers to a failure mode in which fine-tuning large language models on narrowly scoped data induces broadly misaligned behavior.<n>Across multiple domains and model families, we find that fine-tuning models on data exhibiting specific character-level dispositions induces substantially stronger and more transferable misalignment than incorrect-advice fine-tuning.
  arXiv  Detail & Related papers  (2026-01-30T15:28:42Z)
• Benchmarking Corruption Robustness of LVLMs: A Discriminative Benchmark and Robustness Alignment Metric [49.393713730706445]
  We introduce Bench-C, a benchmark emphasizing discriminative samples for assessing corruption robustness.<n>We propose the Robustness Alignment Score (RAS), a unified metric that measures degradation in logit-level prediction structure.
  arXiv  Detail & Related papers  (2025-11-24T12:07:56Z)
• DeLeaker: Dynamic Inference-Time Reweighting For Semantic Leakage Mitigation in Text-to-Image Models [55.30555646945055]
  Text-to-Image (T2I) models are vulnerable to semantic leakage.<n>We introduce DeLeaker, a lightweight approach that mitigates leakage by directly intervening on the model's attention maps.<n>SLIM is the first dataset dedicated to semantic leakage.
  arXiv  Detail & Related papers  (2025-10-16T17:39:21Z)
• CleverCatch: A Knowledge-Guided Weak Supervision Model for Fraud Detection [0.36944296923226316]
  CleverCatch is a knowledge-guided weak supervision model designed to detect fraudulent prescription behaviors.<n>Our approach integrates structured domain expertise into a neural architecture that aligns rules and data samples within a shared embedding space.<n> Experiments on the large-scale real-world dataset demonstrate that CleverCatch outperforms four state-of-the-art anomaly detection baselines.
  arXiv  Detail & Related papers  (2025-10-15T06:49:31Z)
• Understanding Catastrophic Interference: On the Identifibility of Latent Representations [67.05452287233122]
  Catastrophic interference, also known as catastrophic forgetting, is a fundamental challenge in machine learning.<n>We propose a novel theoretical framework that formulates catastrophic interference as an identification problem.<n>Our approach provides both theoretical guarantees and practical performance improvements across both synthetic and benchmark datasets.
  arXiv  Detail & Related papers  (2025-09-27T00:53:32Z)
• Mutual Information Tracks Policy Coherence in Reinforcement Learning [0.0]
  Reinforcement Learning (RL) agents face degradation from sensor faults, actuator wear, and environmental shifts.<n>We present an information-theoretic framework that reveals both the fundamental dynamics of RL and provides practical methods for diagnosing deployment-time anomalies.
  arXiv  Detail & Related papers  (2025-09-12T17:24:20Z)
• Deep Learning Models for Robust Facial Liveness Detection [56.08694048252482]
  This study introduces a robust solution through novel deep learning models addressing the deficiencies in contemporary anti-spoofing techniques.<n>By innovatively integrating texture analysis and reflective properties associated with genuine human traits, our models distinguish authentic presence from replicas with remarkable precision.
  arXiv  Detail & Related papers  (2025-08-12T17:19:20Z)
• Unmasking Synthetic Realities in Generative AI: A Comprehensive Review of Adversarially Robust Deepfake Detection Systems [4.359154048799454]
  Deepfake proliferation-synthetic media poses challenges to digital security, misinformation mitigation, and identity preservation.<n>This systematic review evaluates state-of-the-art deepfake detection methodologies, emphasizing reproducible implementations for transparency and validation.<n>We delineate two core paradigms: (1) detection of fully synthetic media leveraging statistical anomalies and hierarchical feature extraction, and (2) localization of manipulated regions within authentic content employing multi-modal cues such as visual artifacts and temporal inconsistencies.
  arXiv  Detail & Related papers  (2025-07-24T22:05:52Z)
• SafetyNet: Detecting Harmful Outputs in LLMs by Modeling and Monitoring Deceptive Behaviors [2.07180164747172]
  High-risk industries like nuclear and aviation use real-time monitoring to detect dangerous system conditions.<n>We propose a real-time framework to predict harmful AI outputs before they occur by using an unsupervised approach.
  arXiv  Detail & Related papers  (2025-05-20T12:49:58Z)
• Ensuring Medical AI Safety: Interpretability-Driven Detection and Mitigation of Spurious Model Behavior and Associated Data [14.991686165405959]
  We show the applicability of the framework using four medical datasets across two modalities.<n>We successfully identify and unlearn these biases in VGG16, ResNet50, and contemporary Vision Transformer models.
  arXiv  Detail & Related papers  (2025-01-23T16:39:09Z)
• Towards Unbiased Visual Emotion Recognition via Causal Intervention [63.74095927462]
  We propose a novel Emotion Recognition Network (IERN) to alleviate the negative effects brought by the dataset bias. A series of designed tests validate the effectiveness of IERN, and experiments on three emotion benchmarks demonstrate that IERN outperforms other state-of-the-art approaches.
  arXiv  Detail & Related papers  (2021-07-26T10:40:59Z)
• No Need to Know Physics: Resilience of Process-based Model-free Anomaly Detection for Industrial Control Systems [95.54151664013011]
  We present a novel framework to generate adversarial spoofing signals that violate physical properties of the system. We analyze four anomaly detectors published at top security conferences.
  arXiv  Detail & Related papers  (2020-12-07T11:02:44Z)

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.