Metraipolis › archive › dc_aa1ffb6fbd3b
SpanCalib-VLM: Calibrated Hallucination Span Detection in Vision-Language Models — arXiv:2608.29974
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.29974
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T08:08:03.707Z
- SHA-256 of the archived text
172b831ddb35e4e29c625a95dfc70d1c7bb05991437b8ee644902e987d15e4e4- Stable URL of this page
https://metraipolis.com/documento/dc_aa1ffb6fbd3b
How to cite
SpanCalib-VLM: Calibrated Hallucination Span Detection in Vision-Language Models — arXiv:2608.29974 — arXiv (en), https://arxiv.org/abs/2608.29974 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_aa1ffb6fbd3b (Metraipolis archive dc_aa1ffb6fbd3b, sha256 172b831ddb35e4e29c625a95dfc70d1c7bb05991437b8ee644902e987d15e4e4)
BibTeX
@misc{dcaa1ffb6fbd3b,
title = {SpanCalib-VLM: Calibrated Hallucination Span Detection in Vision-Language Models — arXiv:2608.29974},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.29974},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_aa1ffb6fbd3b; sha256 172b831ddb35e4e29c625a95dfc70d1c7bb05991437b8ee644902e987d15e4e4},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.29974. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Detecting hallucinations in Large Vision-Language Models (LVLMs) requires both accurate span localization and well-calibrated confidence scores. Fine-tuned generative VLMs excel at identifying hallucinated text spans but suffer from overconfidence and high inference latency. Discriminative sequence taggers offer deterministic speed and superior calibration but exhibit conservative span recall. We present SpanCalib-VLM, a hybrid dual-system for the SHROOM-Visions Shared Task that combines a multimodal sequence tagger, consisting of XLM-RoBERTa-Large fused with a SigLIP vision encoder via cross-attention, with our fine-tuned generative VLM (Qwen3.5-4B-SHROOM-SFT). Through a Union-Calibrated Fusion strategy, candidate spans from the generative model are re-scored with calibrated probabilities from the sequence tagger. On the SHROOM-Visions English evaluation split, our ensemble achieves a Pearson calibration correlation of 0.41 and an overall IoU of 0.39, with a clean-response IoU of 0.91} and overall detection accuracy of 70.7%. We make our model weights and code publicly available. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.CL). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2608.29974 — 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 2026-10-17T07:47:52.915Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).