Metraipolis › archive › dc_e1a1d8a2f36e

Test-Time Hallucination Control in Large Vision-Language Models — arXiv:2608.11474

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.11474
License (as archived)
CC0-1.0
Archived
2026-10-19T05:49:20.938Z
SHA-256 of the archived text
f597fed492d6cfad7cfee41785b566d016bd8f5df0b1d1fce81e426d435a4a04
Stable URL of this page
https://metraipolis.com/documento/dc_e1a1d8a2f36e

How to cite

Test-Time Hallucination Control in Large Vision-Language Models — arXiv:2608.11474 — arXiv (en), https://arxiv.org/abs/2608.11474 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_e1a1d8a2f36e (Metraipolis archive dc_e1a1d8a2f36e, sha256 f597fed492d6cfad7cfee41785b566d016bd8f5df0b1d1fce81e426d435a4a04)
BibTeX
@misc{dce1a1d8a2f36e,
  title = {Test-Time Hallucination Control in Large Vision-Language Models — arXiv:2608.11474},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.11474},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_e1a1d8a2f36e; sha256 f597fed492d6cfad7cfee41785b566d016bd8f5df0b1d1fce81e426d435a4a04},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.11474. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Object Hallucination in large vision-language models (LVLMs), where models generate non-factual content about input images, remains a critical barrier to their reliability in real-world applications. Existing mitigation strategies can be categorized into training-based and training-free methods. Training-based methods often achieve strong performance but are costly, requiring extensive computational resources, large-scale data, and time-consuming fine-tuning. Training-free approaches are particularly appealing due to their efficiency. However, existing training-free methods either require multiple decoding rounds, which adds computational overhead, or modify internal states in a model-specific way that risks degrading pretrained knowledge. We propose Test-Time Hallucination Mitigation (TTH) method, a novel training-free method that addresses both limitations. TTH introduces a token-validator module, implemented as a zero-shot Multi-Modal Classifier (MMC), to generate auxiliary logits grounded in the input image. These logits are fused with the original LVLM outputs at the token level for object tokens selected from a candidate pool. An entropy-based weighting scheme is then applied to enable robust and accurate predictions. Extensive experiments across multiple LVLM families and diverse benchmarks demonstrate that TTH consistently improves accuracy and robustness, underscoring its generalizability and practical effectiveness. Code is released at this https URL

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-08-13 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2608.11474 — 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 2026-10-19T05:42:25.863Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-08-13?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).