Metraipolis › archive › dc_372bfd3a9e2f
Hallucination Mitigation for Large Vision-Language Models via Implicit Feature Stabilization — arXiv:2608.29924
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2608.29924
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T07:52:18.359Z
- SHA-256 of the archived text
e6c36b1de2a39a087827987c865d5dfeb26feab733b275a1c2d0add980782afc- Stable URL of this page
https://metraipolis.com/documento/dc_372bfd3a9e2f
How to cite
Hallucination Mitigation for Large Vision-Language Models via Implicit Feature Stabilization — arXiv:2608.29924 — arXiv (en), https://arxiv.org/abs/2608.29924 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_372bfd3a9e2f (Metraipolis archive dc_372bfd3a9e2f, sha256 e6c36b1de2a39a087827987c865d5dfeb26feab733b275a1c2d0add980782afc)
BibTeX
@misc{dc372bfd3a9e2f,
title = {Hallucination Mitigation for Large Vision-Language Models via Implicit Feature Stabilization — arXiv:2608.29924},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2608.29924},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_372bfd3a9e2f; sha256 e6c36b1de2a39a087827987c865d5dfeb26feab733b275a1c2d0add980782afc},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.29924. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Large Vision-Language Models (LVLMs) are prone to hallucinations: they fluently describe objects, attributes, and scenes that are not in the image. We connect part of this failure to a measurable property of their representations, feature instability, where mild semantics-preserving perturbations of the input cause large changes in the learned embeddings; hallucination rates rise together with this variability. Existing stability-motivated remedies are explicit, in the sense that they intervene at inference time through latent steering or constrained decoding, and pay for it on every query. We propose implicit stabilization instead: perturbation-invariance is built into the model weights during fine-tuning, and nothing extra runs at deployment. Our framework, INFUSE, first stabilizes visual and textual representations around perturbation-averaged and ground-truth anchors, then aligns the stabilized representations across modalities with bidirectional contrastive objectives. We prove that the anchor's root-mean-square deviation from the perturbation-mean representation shrinks at rate $1/\sqrt{K}$ in the number of views, and that under a Lipschitz decoder, this bounds how much any perturbation can change the model's hallucination behavior. On LLaVA-1.5, LLaVA-1.6, and Qwen3-VL-8B-Instruct, INFUSE reduces AMBER CHAIR by 46-63% relative to each base model, improves ObjHal, MMHal, HallusionBench, and POPE, and preserves VQA-v2 and TextVQA, all with no inference-time overhead.
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.AI).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.29924 — 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 2026-10-17T07:28:26.805Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).