Across four multimodal LLM backbones, MRPO consistently outperforms standard GRPO and a recent RL baseline, and on Qwen3-VL-8B-Thinking even surpasses substantially larger medical MLLMs such as HuatuoGPT-Vision-34B by 4.59 points.
Causal interventions further show that predictions remain sensitive to recurrent content when the question is held fixed, and that persistent visual evidence causally revises the recurrent trajectory.
Suhyeong Park, Jun-Ha Jung, Jaewoo Kang· 1 citation
This work hypothesizes and empirically verify that a model genuinely relies on visual evidence only when its multimodal distribution significantly diverges from its text-only distribution and proposes BUZZY, a training-free decoding method that corrects multimodal predictions by subtracting the text-only distribution.
Taeyun Roh, Suhyeong Park, Dongyoung Lee et al.· 0 citations
SaMer is proposed, an object-aware token merging framework that compresses image-side post-projector tokens into representative centroids while preserving the original late-interaction interface, and outperforms compression baselines and shows stronger phrase-level grounding, suggesting that efficient multi-vector retr...
Suhyeong Park, Junha Jung, Jungwoo Park et al.· arXiv.org· 1 citation
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
