Metraipolis › archive › dc_4acced095257
DreamRealm: Exploring and Materializing Hallucination Worlds for Trustworthy Vision-Language Models — arXiv:2611.06433
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.06433
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-05T18:30:04.598Z
- SHA-256 of the archived text
0e60e60cf9653e749b6ca2824a12c7890be0cc7c3936a762eb8b1eb5f5935587- Stable URL of this page
https://metraipolis.com/documento/dc_4acced095257
How to cite
DreamRealm: Exploring and Materializing Hallucination Worlds for Trustworthy Vision-Language Models — arXiv:2611.06433 — arXiv (en), https://arxiv.org/abs/2611.06433 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4acced095257 (Metraipolis archive dc_4acced095257, sha256 0e60e60cf9653e749b6ca2824a12c7890be0cc7c3936a762eb8b1eb5f5935587)
BibTeX
@misc{dc4acced095257,
title = {DreamRealm: Exploring and Materializing Hallucination Worlds for Trustworthy Vision-Language Models — arXiv:2611.06433},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.06433},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4acced095257; sha256 0e60e60cf9653e749b6ca2824a12c7890be0cc7c3936a762eb8b1eb5f5935587},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.06433. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multimodal large language models (MLLMs) remain prone to visual hallucinations when correct responses depend on subtle visual details. One line of post-training methods constructs visual contrasts through image retrieval, generation, or editing to improve fine-grained perception. However, these contrasts are not built around the model's own hallucinations, making it difficult to ensure that supervision targets its actual perceptual confusions. Another line suppresses hallucinated outputs through response correction or preference learning under fixed image conditions, but leaves the design of supervision signals on the image-input side underexplored. Inspired by how humans correct misconceptions, we introduce DreamRealm, a post-training framework that explores diverse latent hallucinated responses, minimally edits original images to make them valid, and trains the model to assign each response to its supporting image. Our data construction pipeline, RealmForgePipe, mines observed and latent hallucinations from a target model and materializes them through image editing. Using hallucinations mined from Qwen3.5-9B, we construct the Realized Hallucination Dataset (RealmSet). Our training objective, RealmMatch, constructs a likelihood score matrix across image--response combinations and combines supervised learning with bidirectional contrastive losses to reinforce correct matches and suppress mismatches. Compared with the original Qwen3.5-9B model, DreamRealm improves performance across six visual perception and hallucination benchmarks, including gains of 7.26 percentage points in HallusionBench accuracy and 11.34 percentage points in MMVP pair accuracy. Together with ablation studies on data construction and training methods, these results demonstrate that hallucination assignment is a promising paradigm for improving visual perception and mitigating hallucinations. Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.CV, nuovi lavori del giorno). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2611.06433 — 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 2026-11-05T18:19:49.058Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).