Neural Uncertainty Principle: A Unified View of Adversarial Fragility and LLM Hallucination

2026-03-19

AI Summary

This work addresses the prevailing treatment of adversarial vulnerability in vision models and hallucination in large language models as isolated phenomena, lacking a unified theoretical foundation. The authors propose the Neural Uncertainty Principle (NUP), which establishes that an input and its loss gradient—viewed as conjugate observables—are fundamentally constrained by an irreducible uncertainty bound. This principle provides the first unified theoretical framework linking these two challenges. Building on NUP, they introduce ConjMask, a training-free masking strategy that enhances visual robustness without adversarial training, and develop a single backward pass probe during prefilling combined with LogitReg logit regularization to detect hallucination risk prior to generation. Their approach requires neither additional training nor modifications to the decoding process, enabling informed prompt selection and cross-modal reliability assessment.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

Adversarial vulnerability in vision and hallucination in large language models are conventionally viewed as separate problems, each addressed with modality-specific patches. This study first reveals that they share a common geometric origin: the input and its loss gradient are conjugate observables subject to an irreducible uncertainty bound. Formalizing a Neural Uncertainty Principle (NUP) under a loss-induced state, we find that in near-bound regimes, further compression must be accompanied by increased sensitivity dispersion (adversarial fragility), while weak prompt-gradient coupling leaves generation under-constrained (hallucination). Crucially, this bound is modulated by an input-gradient correlation channel, captured by a specifically designed single-backward probe. In vision, masking highly coupled components improves robustness without costly adversarial training; in language, the same prefill-stage probe detects hallucination risk before generating any answer tokens. NUP thus turns two seemingly separate failure taxonomies into a shared uncertainty-budget view and provides a principled lens for reliability analysis. Guided by this NUP theory, we propose ConjMask (masking high-contribution input components) and LogitReg (logit-side regularization) to improve robustness without adversarial training, and use the probe as a decoding-free risk signal for LLMs, enabling hallucination detection and prompt selection. NUP thus provides a unified, practical framework for diagnosing and mitigating boundary anomalies across perception and generation tasks.

Problem

• adversarial fragility
• LLM hallucination
• neural uncertainty
• input-gradient correlation
• model reliability

Innovation

• Neural Uncertainty Principle
• adversarial fragility
• LLM hallucination
• input-gradient correlation
• uncertainty bound