Metraipolis › archive › dc_503c400bd0ea

MergeHEIR: Mitigating Multimodal Hallucinations as the Tax of Model Merging — arXiv:2609.32422

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.32422
License (as archived)
CC0-1.0
Archived
2026-10-07T20:55:25.672Z
SHA-256 of the archived text
65ee95c812cc856866cd7d87731e41d820614a50c9a57fd5fda58c7632b3d849
Stable URL of this page
https://metraipolis.com/documento/dc_503c400bd0ea

How to cite

MergeHEIR: Mitigating Multimodal Hallucinations as the Tax of Model Merging — arXiv:2609.32422 — arXiv (en), https://arxiv.org/abs/2609.32422 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_503c400bd0ea (Metraipolis archive dc_503c400bd0ea, sha256 65ee95c812cc856866cd7d87731e41d820614a50c9a57fd5fda58c7632b3d849)
BibTeX
@misc{dc503c400bd0ea,
  title = {MergeHEIR: Mitigating Multimodal Hallucinations as the Tax of Model Merging — arXiv:2609.32422},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.32422},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_503c400bd0ea; sha256 65ee95c812cc856866cd7d87731e41d820614a50c9a57fd5fda58c7632b3d849},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.32422. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Model merging consolidates task-specialized experts into a single deployable model. However, we show that such capability consolidation incurs a merging tax of increased hallucination: across 8 model-merging methods, every merged checkpoint exhibits a higher hallucination rate than the average of its constituent experts. An intuitive approach is to adapt existing hallucination-mitigation methods to the post-merge model, yet this unconstrained adaptation disrupts inherited capabilities, creating a tension between hallucination mitigation and expertise retention. To tackle this challenge, we introduce MergeHEIR, a post-merge adaptation framework designed to reduce this merging tax while preserving expertise inherited from initial experts. Using small expert-task calibration sets, MergeHEIR constructs layer-wise null-space projectors via SVD from task-specific activations collected from the merged checkpoint, and periodically projects the accumulated post-merge displacement onto the resulting null spaces to preserve inherited expertise. Theoretically, we establish minimum-distortion and maximum-dimensionality guarantees, characterize the threshold-controlled adaptation-retention trade-off, and extend perturbation guarantees beyond finite calibration data. Across 24 paired comparisons spanning three MLLM configurations and 8 model-merging methods, MergeHEIR consistently mitigates hallucination while largely preserving inherited expertise, demonstrating a more favorable hallucination-retention trade-off.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.32422 — 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 2026-10-07T20:29:00.788Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).