mlsanityData to 14 Oct 2026

Two-Token Features and Small-Large Ensembles for VLM Hallucination Detection

arXiv 2609.10244 ↗First posted 9 Sep 2026cs.CLcited by 0

Not on the map. Why?

Abstract (arXiv)

We present our system for the SHROOM-Visions 2026 shared task on character-level VLM hallucination detection. A small (4B-parameter) VLM is fine-tuned as a per-token classifier reading a two-token feature from its own hidden states, and is ensembled with a ~400B zero-shot VLM judge at prediction time. Both components see off-the-shelf OCR of any visible in-image text. We use synthetic hallucination data generated by the large model as a source of ensemble diversity, and use validation to select feature layer, training data and OCR grounding. Our official entry reaches mean Cor 0.487 / Cor-lbl 0.387 on the hidden test set, placing 6th/28 (EN), 6th/21 (FR), 8th/21 (IT) and 7th/22 (ZH) on the task's primary Cor-lbl metric.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 4 of its 10 references that are in the map. First posted on arXiv 9 Sep 2026.

Builds on

4 of 10 references.

References · 4

  1. 2024INSIDE: LLMs' Internal States Retain the Power of Hallucination Detectioncited by 323
  2. 2020PP-OCR: A Practical Ultra Lightweight OCR Systemcited by 151
  3. 2025SemEval-2025 Task 3: Mu-SHROOM, the Multilingual Shared Task on Hallucinations and Related Observable Overgeneration Mistakescited by 27
  4. 2026Can Humans Dream of Electric Sheep? Human-Written Samples for Fine-Grained Vision-and-Language Hallucination Benchmarkingcited by 4

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025MTRE: Multi-Token Reliability Estimation for Hallucination Detection in VLMs
  2. 2026Beyond the Global Scores: Fine-Grained Token Grounding as a Robust Detector of LVLM Hallucinations
  3. 2026EnsemHalDet: Robust VLM Hallucination Detection via Ensemble of Internal State Detectors
  4. 2026Density Ridge Selective Prediction for LLM and VLM Hallucination Detection under Calibration Label Scarcity
  5. 2026HALP: Detecting Hallucinations in Vision-Language Models without Generating a Single Token
  6. 2025Leveraging NTPs for Efficient Hallucination Detection in VLMs
  7. 2026VisER: Visual Evidence and Reliance for Object Hallucination Detection in LVLMs
  8. 2026UniProbe: A Learnable Token-Level Hallucination Detector for Large VLMs using Multi-Structural Internal Representations