mlsanityData to 14 Oct 2026

Leveraging NTPs for Efficient Hallucination Detection in VLMs

arXiv 2509.20379 ↗First posted 20 Sep 2025cs.CV cs.CL cs.LGcited by 1

On the map: Multimodal LLMs / vision-language models → Hallucination mitigation in LVLMs

Abstract (arXiv)

Hallucinations of vision-language models (VLMs), which are misalignments between visual content and generated text, undermine the reliability of VLMs. One common approach for detecting them employs the same VLM, or a different one, to assess generated outputs. This process is computationally intensive and increases model latency. In this paper, we explore an efficient on-the-fly method for hallucination detection by training traditional ML models over signals based on the VLM's next-token probabilities (NTPs). NTPs provide a direct quantification of model uncertainty. We hypothesize that high uncertainty (i.e., a low NTP value) is strongly associated with hallucinations. To test this, we introduce a dataset of 1,400 human-annotated statements derived from VLM-generated content, each labeled as hallucinated or not, and use it to test our NTP-based lightweight method. Our results demonstrate that NTP-based features are valuable predictors of hallucinations, enabling fast and simple ML models to achieve performance comparable to that of strong VLMs. Furthermore, augmenting these NTPs with linguistic NTPs, computed by feeding only the generated text back into the VLM, enhances hallucination detection performance. Finally, integrating hallucination prediction scores from VLMs into the NTP-based models led to better performance than using either VLMs or NTPs alone. We hope this study paves the way for simple, lightweight solutions that enhance the reliability of VLMs.

Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Hallucination mitigation in LVLMs, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 17 of its 28 references that are in the map. First posted on arXiv 20 Sep 2025.

Builds on

17 of 28 references.

Closest ancestors · 3

Same direction, newest first

  1. 2024AutoHallusion: Automatic Generation of Hallucination Benchmarks for Vision-Language Models16 Jun
  2. 2024IBD: Alleviating Hallucinations in Large Vision-Language Models via Image-Biased Decoding28 Feb
  3. 2023Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding28 Nov

Other references · 14

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2019VisualBERT: A Simple and Performant Baseline for Vision and Languagecited by 1,296
  3. 2023Explainability for Large Language Models: A Surveycited by 333
  4. 2024Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Toolscited by 176
  5. 2023Perspectives on Large Language Models for Relevance Judgmentcited by 85
  6. 2020On hallucinations in tomographic image reconstructioncited by 68
  7. 2024Detecting Hallucinations in Large Language Model Generation: A Token Probability Approachcited by 42
  8. 2023Llamas Know What GPTs Don't Show: Surrogate Models for Confidence Estimationcited by 39
  9. 2024AI Hallucinations: A Misnomer Worth Clarifyingcited by 34
  10. 2023Can Linguistic Knowledge Improve Multimodal Alignment in Vision-Language Pretraining?cited by 21
  11. 2024Machine Translation Hallucination Detection for Low and High Resource Languages using Large Language Modelscited by 13
  12. 2024Large Language Models Cannot Explain Themselvescited by 11
2 more
  1. 2023Temperature-scaling surprisal estimates improve fit to human reading times -- but does it do so for the "right reasons"?cited by 7
  2. 2025Mitigating Hallucinated Translations in Large Language Models with Hallucination-focused Preference Optimizationcited by 6

Built on by

Cited by year

Citing papers per year: 2025: 1.

1 paper.

  1. 2025Silenced Biases: The Dark Side LLMs Learned to Refusecited by 10

Similar, not cited

Close in wording, no citation either way

  1. 2024VL-Uncertainty: Detecting Hallucination in Large Vision-Language Model via Uncertainty Estimation
  2. 2026HALP: Detecting Hallucinations in Vision-Language Models without Generating a Single Token
  3. 2025MTRE: Multi-Token Reliability Estimation for Hallucination Detection in VLMs
  4. 2026UHP Detection: LVLMs have their Unique Hallucination Pattern in the Consistency Space
  5. 2024Reference-free Hallucination Detection for Large Vision-Language Models
  6. 2025Revisiting Hallucination Detection with Effective Rank-based Uncertainty
  7. 2025HalLoc: Token-level Localization of Hallucinations for Vision Language Models
  8. 2026Scalable Token-Level Hallucination Detection in Large Language Models