mlsanityData to 14 Oct 2026

MTRE: Multi-Token Reliability Estimation for Hallucination Detection in VLMs

arXiv 2505.11741 ↗First posted 16 May 2025cs.AI cs.CRcited by 3

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

Abstract (arXiv)

Vision-language models (VLMs) now rival human performance on many multimodal tasks, yet they still hallucinate objects or generate unsafe text. Current hallucination detectors, e.g., single-token linear probing (LP) and PTrue, typically analyze only the logit of the first generated token or just its highest-scoring component, overlooking richer signals embedded within earlier token distributions. We demonstrate that analyzing the complete sequence of early logits potentially provides substantially more diagnostic information. We emphasize that hallucinations may only emerge after several tokens, as subtle inconsistencies accumulate over time. By analyzing the Kullback-Leibler (KL) divergence between logits corresponding to hallucinated and non-hallucinated tokens, we underscore the importance of incorporating later-token logits to more accurately capture the reliability dynamics of VLMs. In response, 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. Despite the challenges posed by large vocabulary sizes and long logit sequences, MTRE remains efficient and tractable. Across MAD-Bench, MM-SafetyBench, MathVista, and four compositional-geometry benchmarks, MTRE achieves a 9.4% gain in accuracy and a 14.8% gain in AUROC over standard detection methods, establishing a new state of the art in hallucination detection for open-source VLMs.

Key facts. Cited by 3 papers 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 25 of its 34 references that are in the map. First posted on arXiv 16 May 2025.

Builds on

25 of 34 references.

Closest ancestors · 2

Same direction, newest first

  1. 2024The First to Know: How Token Distributions Reveal Hidden Knowledge in Large Vision-Language Models?14 Mar
  2. 2024Uncertainty-Aware Evaluation for Vision-Language Models22 Feb

Other references · 23

  1. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,170
  2. 2017On Calibration of Modern Neural Networkscited by 4,960
  3. 2015Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learningcited by 4,448
  4. 2023Improved Baselines with Visual Instruction Tuningcited by 4,418
  5. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,356
  6. 2017What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?cited by 2,352
  7. 2022Language Models (Mostly) Know What They Knowcited by 1,856
  8. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contextscited by 1,614
  9. 2023Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generationcited by 819
  10. 2023LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Modelcited by 602
  11. 2023mPLUG-Owl2: Revolutionizing Multi-modal Large Language Model with Modality Collaborationcited by 547
  12. 2023Language Models Represent Space and Timecited by 316
11 more
  1. 2023Visual-Language Prompt Tuning with Knowledge-guided Context Optimizationcited by 308
  2. 2024LLMs Know More Than They Show: On the Intrinsic Representation of LLM Hallucinationscited by 244
  3. 2023Shifting Attention to Relevance: Towards the Predictive Uncertainty Quantification of Free-Form Large Language Modelscited by 150
  4. 2024Vision language models are blind: Failing to translate detailed visual features into wordscited by 147
  5. 2023Conformal Prediction with Large Language Models for Multi-Choice Question Answeringcited by 139
  6. 2024MLLM-Protector: Ensuring MLLM's Safety without Hurting Performancecited by 117
  7. 2024What Large Language Models Know and What People Think They Knowcited by 110
  8. 2023DRESS: Instructing Large Vision-Language Models to Align and Interact with Humans via Natural Language Feedbackcited by 103
  9. 2023Fine-tuning can cripple your foundation model; preserving features may be the solutioncited by 96
  10. 2022Looking for a Needle in a Haystack: A Comprehensive Study of Hallucinations in Neural Machine Translationcited by 61
  11. 2024Improving Uncertainty Quantification in Large Language Models via Semantic Embeddingscited by 30

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026PARALLAX: Separating Genuine Hallucination Detection from Benchmark Construction Artifactscited by 3
  2. 2026Geometric Risk Control for Vision-Language Model OCRcited by 1
  3. 2026HalluScope: Fine-grained Hallucination Diagnosis for Multimodal Large Language Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026HALP: Detecting Hallucinations in Vision-Language Models without Generating a Single Token
  2. 2026Beyond the Global Scores: Fine-Grained Token Grounding as a Robust Detector of LVLM Hallucinations
  3. 2025Leveraging NTPs for Efficient Hallucination Detection in VLMs
  4. 2025Seeing is Believing: Rich-Context Hallucination Detection for MLLMs via Backward Visual Grounding
  5. 2026Scalable Token-Level Hallucination Detection in Large Language Models
  6. 2026UHP Detection: LVLMs have their Unique Hallucination Pattern in the Consistency Space
  7. 2025A Survey of Multimodal Hallucination Evaluation and Detection
  8. 2024VL-Uncertainty: Detecting Hallucination in Large Vision-Language Model via Uncertainty Estimation