Metraipolis › archive › dc_b03e2e2ee185

UniProbe: A Learnable Token-Level Hallucination Detector for Large VLMs using Multi-Structural Internal Representations — arXiv:2608.10835

Source: arXiv (en)
Original: https://arxiv.org/abs/2608.10835
License (as archived): CC0-1.0
Archived: 2026-10-19T06:19:50.249Z
SHA-256 of the archived text: 557ce714d3e3ccec2f6234320af07bc5749343a7a334b7a56582a6add7bb6b59
Stable URL of this page: https://metraipolis.com/documento/dc_b03e2e2ee185

How to cite

UniProbe: A Learnable Token-Level Hallucination Detector for Large VLMs using Multi-Structural Internal Representations — arXiv:2608.10835 — arXiv (en), https://arxiv.org/abs/2608.10835 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_b03e2e2ee185 (Metraipolis archive dc_b03e2e2ee185, sha256 557ce714d3e3ccec2f6234320af07bc5749343a7a334b7a56582a6add7bb6b59)

BibTeX

@misc{dcb03e2e2ee185,
  title = {UniProbe: A Learnable Token-Level Hallucination Detector for Large VLMs using Multi-Structural Internal Representations — arXiv:2608.10835},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.10835},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_b03e2e2ee185; sha256 557ce714d3e3ccec2f6234320af07bc5749343a7a334b7a56582a6add7bb6b59},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.10835. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Large Vision-Language Models (LVLMs) achieve impressive visual reasoning and dialogue capabilities, yet frequently hallucinate content unsupported by the visual input. Effective mitigation requires token-level localization, enabling targeted intervention without discarding the entire response. Existing detectors require expensive full-model fine-tuning, rely on external verifiers that ignore the model's generation process, or reduce internal signals to isolated features and hand-crafted statistics, discarding spatial, sequential, and relational structure. We introduce \textbf{UniProbe}, a lightweight, unified, learnable detector that models a frozen LVLM's heterogeneous computational trace from a single forward pass. UniProbe constructs a directed graph over image patches, query tokens, and generated tokens, with attention weights encoding their relations. It processes this trace with alternating structure-aware modules: a GNN for relational evidence, a ViT for 2-D visual geometry, and a GRU for response order. Interleaving them allows spatial, relational, and sequential evidence to interact throughout the detector. We further develop a streaming variant for hallucination-aware decoding, which detects and resamples hallucinated tokens during generation, and a self-adaptation strategy aligning the detector with the LVLM's own generations. Across diverse LVLM backbones, UniProbe achieves state-of-the-art token-level and object-hallucination detection. During decoding, it reduces object hallucinations by up to 55\% at $1.06\times$ the latency of standard generation.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-12 in cs.LG).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.10835 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-19T06:11:45.147Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-08-12?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).