Quickest Detection of Hallucination Onset: Delay Bounds and Learned CUSUM Statistics

• URL: http://arxiv.org/abs/2606.12476v2
• Date: Fri, 12 Jun 2026 13:24:39 GMT
• Title: Quickest Detection of Hallucination Onset: Delay Bounds and Learned CUSUM Statistics
• Abstract summary: hallucination onset detection is a quickest change detection problem.<n>A causal recurrent labeler acts as a CUSUM with a learned increment.<n>Low-false-alarm onset detection is hard.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Token-level hallucination detectors are evaluated as classifiers, by AUC over all tokens, yet a streaming monitor is judged by its reaction time: the number of tokens that pass between the onset of a hallucination and the alarm. We formulate hallucination onset detection as a quickest change detection problem. A first-order Markov model of the latent faithful/hallucinated state, validated on RAGTruth, places the task inside classical change-point theory and yields Lorden's lower bound on detection delay: about 1.3 tokens at a false-alarm rate of 0.01. We then show that a causal recurrent labeler acts as a CUSUM with a learned increment. Among the onsets it catches it detects in 11-13 tokens, against 31 for a linear per-token baseline, though at this false-alarm budget every detector catches under a third of onsets and the recall-honest delay is 56-66 tokens: low-false-alarm onset detection is hard. A controlled decomposition attributes the speed advantage mostly to a better per-token score rather than to temporal accumulation. An information-rate optimality theorem of Donsker-Varadhan type explains the remaining order-of-magnitude gap: the learned score realizes only 1/4.5 of the divergence the features carry, a deficit that recalibration cannot remove, with the remainder a finite-horizon effect. Classification metrics conceal this delay structure; sequential analysis makes it measurable.

Related papers

• Not All Relations Are Equal: Relation-Balanced and Calibrated Graph Learning for Provenance-Based Intrusion Detection [66.44527094471619]
  We present RECAL, an unsupervised framework using relation-balanced masked graph learning to better capture rare interaction patterns.<n>On three DARPA E3 datasets, RECAL achieves F1 scores of 99.99%, 99.93%, and 99.99%, outperforming the best baseline on each dataset by 0.88, 0.82, and 0.42 percentage points, respectively.
  arXiv  Detail & Related papers  (2026-09-15T00:31:01Z)
• Beyond Lexical Metrics: Sentence-Embedding Detection of Reviewer Habituation in AI Code Review [2.9152186617858304]
  We study 11,429 reviews paired with 10,104 human-authored inline comments from AI.<n>Approval rates rise from 30.5% in reviewers' early periods to 36.6% in late periods.<n>Four hand-crafted linguistic features - lexical diversity, Shannon entropy, technical specificity, and constructive actionability - show no monotonic decline across exposure deciles.
  arXiv  Detail & Related papers  (2026-09-05T18:15:18Z)
• TriggerBench: Investigating Prospective Memory for Large Language Models [61.96701552522981]
  We introduce TriggerBench, a benchmark for Prospective memory (PM)<n>PM shows a precision-recall trade-off and attentional fragility.<n>PM is notably harder than RM: on identical contexts, near-saturates up to 100K tokens, while PM decays sharply as context length scales.
  arXiv  Detail & Related papers  (2026-06-22T15:07:36Z)
• Amplify, Don't Create: Temporal Accumulation for Slow-Burn Prompt Injection [0.0]
  Control-plane attacks against tool-using agents can instead distribute weak directives across a trajectory while keeping each event below threshold.<n>We test whether a proxy-side temporal accumulator recovers this slow-burn signal by reducing frozen per-event scores to peak and CUSUM persistence statistics.
  arXiv  Detail & Related papers  (2026-06-17T23:39:38Z)
• The Hiremath Early Detection (HED) Score: A Measure-Theoretic Evaluation Standard for Temporal Intelligence [0.0]
  We introduce the Hiremath Early Detection (HED) Score, a principled, measuretheoretic criterion for evaluation of systems operating over nonstationary processes.<n>The HED Score integrates a baseline-neutral, exponentially decaying kernel over the posterior probability stream of a target regime, beginning precisely at the onset of the regime shift.
  arXiv  Detail & Related papers  (2026-04-05T14:22:20Z)
• Neural Uncertainty Principle: A Unified View of Adversarial Fragility and LLM Hallucination [60.197429875410286]
  Adversarial vulnerability in vision and hallucination in large language models are conventionally viewed as separate problems.<n>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.<n>In vision, masking highly coupled components improves robustness without costly adversarial training.<n>In language, the same prefill-stage probe detects hallucination risk before generating any answer tokens.
  arXiv  Detail & Related papers  (2026-03-20T02:07:10Z)
• Early Rug Pull Warning for BSC Meme Tokens via Multi-Granularity Wash-Trading Pattern Profiling [3.1262100364392205]
  High-frequency issuance and short-cycle speculation of meme tokens in decentralized finance (DeFi) have significantly amplified rug-pull risk.<n>Existing approaches still struggle to provide stable early warning under scarce anomalies, incomplete labels, and limited interpretability.<n>To address this issue, an end-to-end warning framework is proposed for BSC meme tokens, consisting of four stages: dataset construction and labeling, wash-trading pattern feature modeling, risk prediction, and error analysis.
  arXiv  Detail & Related papers  (2026-03-14T08:25:42Z)
• LEFT: Learnable Fusion of Tri-view Tokens for Unsupervised Time Series Anomaly Detection [53.191369031661885]
  Unsupervised time series anomaly detection aims to build a model for identifying abnormal timestamps without assuming the availability of annotations.<n>We present Learnable Fusion of Tri-view Tokens (LEFT), a unified unsupervised TSAD framework that models anomalies as inconsistencies across complementary representations.<n>Experiments on real-world benchmarks show that LEFT yields the best detection accuracy against SOTA baselines, while achieving a 5x reduction on FLOPs and 8x speed-up for training.
  arXiv  Detail & Related papers  (2026-02-09T13:33:49Z)
• Probability-Entropy Calibration: An Elastic Indicator for Adaptive Fine-tuning [55.2818264614932]
  RankTuner introduces a probability--entropy calibration signal, the Relative Rank Indicator, which compares the rank of the ground-truth token with its expected rank under the prediction distribution.<n>The inverse indicator is used as a token-wise Relative Scale to reweight the fine-tuning objective, focusing updates on truly under-learned tokens.
  arXiv  Detail & Related papers  (2026-02-02T07:27:19Z)
• Finite-Horizon Quickest Change Detection Balancing Latency with False Alarm Probability [20.918853345531375]
  A finite-horizon variant of the quickest change detection (QCD) problem that is of relevance to learning in non-stationary environments is studied.<n>The metric characterizing false alarms is the probability of a false alarm occurring before the horizon ends.<n>The objective is to minimize the latency (at a given latency level) while maintaining a low false alarm probability.
  arXiv  Detail & Related papers  (2025-11-16T22:16:33Z)
• Quickest Change Detection in Continuous-Time in Presence of a Covert Adversary [7.58317340007754]
  We investigate the problem of covert quickest change detection in a continuous-time setting, where a Brownian motion experiences a drift change at an unknown time.<n>We show that the adversary achieves damage maximal when the drift scales as $mu(gamma) = Theta (1/sqrtgamma)$, marking a fundamental trade-off between stealth and impact in continuous-time detection systems.
  arXiv  Detail & Related papers  (2025-09-22T13:40:53Z)
• Uncertainty-Aware Attention Heads: Efficient Unsupervised Uncertainty Quantification for LLMs [129.79394562739705]
  Large language models (LLMs) exhibit impressive fluency, but often produce critical errors known as "hallucinations"<n>We propose RAUQ (Recurrent Attention-based Uncertainty Quantification), an unsupervised approach that leverages intrinsic attention patterns in transformers to detect hallucinations efficiently.<n> Experiments across 4 LLMs and 12 question answering, summarization, and translation tasks demonstrate that RAUQ yields excellent results.
  arXiv  Detail & Related papers  (2025-05-26T14:28:37Z)
• MTRE: Multi-Token Reliability Estimation for Hallucination Detection in VLMs [2.6076093883457454]
  Vision-language models (VLMs) now rival human performance on many multimodal tasks, yet they still hallucinate objects or generate unsafe text.<n>We demonstrate that analyzing the complete sequence of early logits potentially provides substantially more diagnostic information.<n>We introduce Multi-Token Reliability Estimation (MTRE), a lightweight, white-box method that aggregates logits from the first ten tokens using multi-token log-likelihood ratios and self-attention.
  arXiv  Detail & Related papers  (2025-05-16T23:00:19Z)
• CARLA: Self-supervised Contrastive Representation Learning for Time Series Anomaly Detection [53.83593870825628]
  One main challenge in time series anomaly detection (TSAD) is the lack of labelled data in many real-life scenarios. Most of the existing anomaly detection methods focus on learning the normal behaviour of unlabelled time series in an unsupervised manner. We introduce a novel end-to-end self-supervised ContrAstive Representation Learning approach for time series anomaly detection.
  arXiv  Detail & Related papers  (2023-08-18T04:45:56Z)
• Optimal Sequential Detection of Signals with Unknown Appearance and Disappearance Points in Time [64.26593350748401]
  The paper addresses a sequential changepoint detection problem, assuming that the duration of change may be finite and unknown. We focus on a reliable maximin change detection criterion of maximizing the minimal probability of detection in a given time (or space) window. The FMA algorithm is applied to detecting faint streaks of satellites in optical images.
  arXiv  Detail & Related papers  (2021-02-02T04:58:57Z)

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.