Text Corpora as Concept Fields: Black-Box Hallucination and Novelty Measurement

• URL: http://arxiv.org/abs/2605.05103v2
• Date: Sat, 09 May 2026 17:34:20 GMT
• Title: Text Corpora as Concept Fields: Black-Box Hallucination and Novelty Measurement
• Abstract summary: We introduce the textbfConcept Field of a text corpus: a local drift field with pointwise uncertainty.<n>We score its agreement with the field by $$, the mean absolute z-distance between the observed delta and the field's local Gaussian estimate.<n>We evaluate this approach on two large-scale settings: groundedness detection over the U.S. Code of Federal Regulations, and novelty detection over Project Gutenberg.
• Score: 3.035872403903575
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We introduce the \textbf{Concept Field} of a text corpus: a local drift field with pointwise uncertainty, estimated in sentence-embedding space from the deltas between consecutive sentences. Given a candidate sentence transition, we score its agreement with the field by $ζ$, the mean absolute z-distance between the observed delta and the field's local Gaussian estimate. The score is black-box (no model internals), corpus-attributable (every score traces to nearby corpus sentences), and admits a probabilistically motivated interpretation under a local Gaussian approximation. We support the computation with the introduction of a \textbf{Vector Sequence Database (VSDB)} that stores embeddings together with sequence-position and next-delta metadata. We evaluate this approach on two large-scale settings: hallucination-style groundedness detection over the U.S. Code of Federal Regulations, and novelty detection over Project Gutenberg. On controlled LLM-generated rewrites, Concept Fields achieve strong selective classification performance under a grounded / ungrounded / unsure triage policy. Unlike retrieval-centric baselines, the resulting coverage-risk behavior is similar across both domains, supporting a degree of cross-domain stability for the standardized deviation score. We also sketch how divergence and curl of the Concept Field, computed on dense clusters, surface qualitatively meaningful semantic patterns (logic sources, sinks, and implicit topics), which we offer as hypothesis-generating rather than as a quantitative result. Concept Fields provide a fast, lightweight, and interpretable signal for groundedness and novelty, complementary to LLM-as-judge and white-box detectors.

Related papers

• Structuring Semantic Embeddings for Principle Evaluation: A Prototype-Guided Contrastive Learning Approach [58.15427952222424]
  General-purpose text embeddings are widely deployed for such tasks, but broad semantic similarity can place semantically similar yet task-distinct examples.<n>We introduce Prototype-Guided Contrastive Learning (PGCL), a prototype-guided geometric regularization module built on top of frozen text embeddings.
  arXiv  Detail & Related papers  (2026-08-15T13:19:25Z)
• Reference-Induced Consensus for Selective Posed-Reference Visual Localization [11.166886974945806]
  RIC-Loc is a scene-training-free posed-reference localizer.<n>It is SfM-point-map-free in its main estimator.
  arXiv  Detail & Related papers  (2026-07-06T06:48:34Z)
• When Do Local Score Models Extrapolate Across Size? A Diagnostic Theory and Benchmark [0.0]
  We show that architectural locality alone does not guarantee stable size extrapolation.<n>We formalize this mechanism, proving a size-uniform comparison theorem for local marginals.<n> Empirically, we validate the interplay between spatial mixing, smoothed-score quasi-locality, and model receptive fields.
  arXiv  Detail & Related papers  (2026-06-08T16:24:31Z)
• CIEC: Coupling Implicit and Explicit Cues for Multimodal Weakly Supervised Manipulation Localization [25.78477436147408]
  Coupling Implicit and Explicit Cues (CIEC) aims to achieve multimodal weakly-supervised manipulation localization for image-text pairs.<n>It integrates forgery cues from both visual and textual perspectives to lock onto suspicious regions aided by spatial priors.<n>For the latter, we devise the Visual-deviation Calibrated Token Grounding (VCTG) module. It focuses on meaningful content words and leverages relative visual bias to assist token localization.
  arXiv  Detail & Related papers  (2026-02-02T14:46:38Z)
• TOPol: Capturing and Explaining Multidimensional Semantic Polarity Fields and Vectors [0.0]
  This work introduces TOPol, a framework for reconstructing and interpreting multidimensional narrative polarity fields.<n>The framework embeds documents using a transformer-based large language model (tLLM)<n>We evaluate TOPol on two corpora: (i) U.S. Central Bank speeches around a macroeconomic breakpoint, and (ii) Amazon product reviews across rating strata.
  arXiv  Detail & Related papers  (2025-10-29T01:14:21Z)
• 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)
• CPR++: Object Localization via Single Coarse Point Supervision [55.8671776333499]
  coarse point refinement (CPR) is first attempt to alleviate semantic variance from an algorithmic perspective. CPR reduces semantic variance by selecting a semantic centre point in a neighbourhood region to replace the initial annotated point. CPR++ can obtain scale information and further reduce the semantic variance in a global region.
  arXiv  Detail & Related papers  (2024-01-30T17:38:48Z)
• Divide and Contrast: Source-free Domain Adaptation via Adaptive Contrastive Learning [122.62311703151215]
  Divide and Contrast (DaC) aims to connect the good ends of both worlds while bypassing their limitations. DaC divides the target data into source-like and target-specific samples, where either group of samples is treated with tailored goals. We further align the source-like domain with the target-specific samples using a memory bank-based Maximum Mean Discrepancy (MMD) loss to reduce the distribution mismatch.
  arXiv  Detail & Related papers  (2022-11-12T09:21:49Z)
• Contextual information integration for stance detection via cross-attention [59.662413798388485]
  Stance detection deals with identifying an author's stance towards a target. Most existing stance detection models are limited because they do not consider relevant contextual information. We propose an approach to integrate contextual information as text.
  arXiv  Detail & Related papers  (2022-11-03T15:04:29Z)
• Concrete Score Matching: Generalized Score Matching for Discrete Data [109.12439278055213]
  "Concrete score" is a generalization of the (Stein) score for discrete settings. "Concrete Score Matching" is a framework to learn such scores from samples.
  arXiv  Detail & Related papers  (2022-11-02T00:41:37Z)
• Contextualized Semantic Distance between Highly Overlapped Texts [85.1541170468617]
  Overlapping frequently occurs in paired texts in natural language processing tasks like text editing and semantic similarity evaluation. This paper aims to address the issue with a mask-and-predict strategy. We take the words in the longest common sequence as neighboring words and use masked language modeling (MLM) to predict the distributions on their positions. Experiments on Semantic Textual Similarity show NDD to be more sensitive to various semantic differences, especially on highly overlapped paired texts.
  arXiv  Detail & Related papers  (2021-10-04T03:59:15Z)

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.