Metraipolis › archive › dc_0eff1d39fa93
From Reward Signal to Visual Utility: A Controlled Audit of Medical VLM Post-Training — arXiv:2609.31450
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.31450
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T01:39:57.616Z
- SHA-256 of the archived text
533caaf519eae94b96759249c8868bfa6a307594421d93532b9cc665e3b18d50- Stable URL of this page
https://metraipolis.com/documento/dc_0eff1d39fa93
How to cite
From Reward Signal to Visual Utility: A Controlled Audit of Medical VLM Post-Training — arXiv:2609.31450 — arXiv (en), https://arxiv.org/abs/2609.31450 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0eff1d39fa93 (Metraipolis archive dc_0eff1d39fa93, sha256 533caaf519eae94b96759249c8868bfa6a307594421d93532b9cc665e3b18d50)
BibTeX
@misc{dc0eff1d39fa93,
title = {From Reward Signal to Visual Utility: A Controlled Audit of Medical VLM Post-Training — arXiv:2609.31450},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.31450},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0eff1d39fa93; sha256 533caaf519eae94b96759249c8868bfa6a307594421d93532b9cc665e3b18d50},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.31450. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Medical vision-language model (VLM) post-training is commonly evaluated through answer accuracy. We examine how changes in accuracy and training objectives relate to image-conditioned decisions in a controlled Qwen2.5-VL-3B study on PMC-VQA. We compare supervised fine-tuning (SFT) with low-rank adaptation (LoRA) restricted to the language model, expanded multimodal adaptation scopes, standard answer-only Group Relative Policy Optimization (GRPO), and a counterfactual evidence objective. On 2,000 clean-test questions, language model LoRA SFT changes correct-image accuracy by +1.10 percentage points (95% paired bootstrap CI:-0.85 to +3.05), while visual-benefit events decrease by 2.40 points and image sensitivity decreases by 5.60 points. Paired records reveal 155 acquired and 203 lost visual-benefit events. Broader adaptation yields lower correct-image accuracy than language-model LoRA SFT. Standard GRPO produces mixed-reward groups and parameter updates, with an uncertain clean test accuracy change. A generation audit reveals that canonical option scores can follow a different token path from generated answers. With scores taken along the greedy generation path, the evidence target improves on the training set; its gains over standard GRPO remain inconsistent on validation data at matched training doses. Sample-level analyses trace how evidence scores, decision margins, and generated answers change during post-training. This empirical and measurement audit identifies gaps between optimization activity, target acquisition, and useful held-out visual behavior. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-28 in cs.AI). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.31450 — 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 2026-10-08T01:12:14.599Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).