Saved in:
Bibliographic Details
Main Authors: Hagen, Luca, Müller, Johanna P., Zhang, Weitong, Qiao, Mengyun, Kainz, Bernhard
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.18313
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913141110079488
author Hagen, Luca
Müller, Johanna P.
Zhang, Weitong
Qiao, Mengyun
Kainz, Bernhard
author_facet Hagen, Luca
Müller, Johanna P.
Zhang, Weitong
Qiao, Mengyun
Kainz, Bernhard
contents Small vision-language models (2-8B) are well-suited for clin- ical deployment due to privacy constraints, limited connectivity, and low-latency requirements favouring on-device or on-premise inference. However, their limited capacity exacerbates the generation of plausible but incorrect outputs. We extend game-theoretic decoding, previously restricted to text-only, closed-ended NLP tasks, to vision-language mod- els for open-ended Medical VQA. We introduce a semantically aware Wasserstein stopping criterion that replaces lexical order matching, en- abling convergence based on semantic consensus among near-synonymous candidate answers and avoiding unnecessary iterations caused by clini- cally equivalent ranking swaps. On VQA-RAD and PathVQA, we ob- tain consistent, statistically significant improvements over greedy and discriminative baselines. On VQA-RAD, we improve Qwen3-VL-2B by +3.5 percentage points (p < 0.01), surpassing the greedy 4B model, with similar trends at larger scales. On PathVQA, Gemma-3-4B with BDG matches MedGemma-4B under greedy decoding despite no domain- specific fine-tuning. At accuracy parity with classic BDG, the Wasser- stein criterion reduces average convergence iterations by approximately 20%, improving inference efficiency while preserving the game-theoretic equilibrium behaviour. Code is available at https://github.com/luca-hagen/ Wasserstein-BDG-medical-VQA.
format Preprint
id arxiv_https___arxiv_org_abs_2605_18313
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Wasserstein Equilibrium Decoding for Reliable Medical Visual Question Answering
Hagen, Luca
Müller, Johanna P.
Zhang, Weitong
Qiao, Mengyun
Kainz, Bernhard
Computer Vision and Pattern Recognition
Artificial Intelligence
Small vision-language models (2-8B) are well-suited for clin- ical deployment due to privacy constraints, limited connectivity, and low-latency requirements favouring on-device or on-premise inference. However, their limited capacity exacerbates the generation of plausible but incorrect outputs. We extend game-theoretic decoding, previously restricted to text-only, closed-ended NLP tasks, to vision-language mod- els for open-ended Medical VQA. We introduce a semantically aware Wasserstein stopping criterion that replaces lexical order matching, en- abling convergence based on semantic consensus among near-synonymous candidate answers and avoiding unnecessary iterations caused by clini- cally equivalent ranking swaps. On VQA-RAD and PathVQA, we ob- tain consistent, statistically significant improvements over greedy and discriminative baselines. On VQA-RAD, we improve Qwen3-VL-2B by +3.5 percentage points (p < 0.01), surpassing the greedy 4B model, with similar trends at larger scales. On PathVQA, Gemma-3-4B with BDG matches MedGemma-4B under greedy decoding despite no domain- specific fine-tuning. At accuracy parity with classic BDG, the Wasser- stein criterion reduces average convergence iterations by approximately 20%, improving inference efficiency while preserving the game-theoretic equilibrium behaviour. Code is available at https://github.com/luca-hagen/ Wasserstein-BDG-medical-VQA.
title Wasserstein Equilibrium Decoding for Reliable Medical Visual Question Answering
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2605.18313