Salvato in:
Dettagli Bibliografici
Autori principali: Kumar, Bhavesh, Feng, Dylan, Tang, Leonard
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:https://arxiv.org/abs/2603.07990
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911540231274496
author Kumar, Bhavesh
Feng, Dylan
Tang, Leonard
author_facet Kumar, Bhavesh
Feng, Dylan
Tang, Leonard
contents Multimodal judges struggle to ground decisions in visual evidence. We present MJ1, a multimodal judge trained with reinforcement learning that enforces visual grounding through a structured grounded verification chain (observations $\rightarrow$ claims $\rightarrow$ verification $\rightarrow$ evaluation $\rightarrow$ scoring) and a counterfactual consistency reward that penalizes position bias. Even without training, our mechanism improves base-model accuracy on MMRB2 by +3.8 points on Image Editing and +1.7 on Multimodal Reasoning. After training, MJ1, with only 3B active parameters, achieves 77.0% accuracy on MMRB2 and surpasses orders-of-magnitude larger models like Gemini-3-Pro. These results show that grounded verification and consistency-based training substantially improve multimodal judgment without increasing model scale.
format Preprint
id arxiv_https___arxiv_org_abs_2603_07990
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle MJ1: Multimodal Judgment via Grounded Verification
Kumar, Bhavesh
Feng, Dylan
Tang, Leonard
Machine Learning
Multimodal judges struggle to ground decisions in visual evidence. We present MJ1, a multimodal judge trained with reinforcement learning that enforces visual grounding through a structured grounded verification chain (observations $\rightarrow$ claims $\rightarrow$ verification $\rightarrow$ evaluation $\rightarrow$ scoring) and a counterfactual consistency reward that penalizes position bias. Even without training, our mechanism improves base-model accuracy on MMRB2 by +3.8 points on Image Editing and +1.7 on Multimodal Reasoning. After training, MJ1, with only 3B active parameters, achieves 77.0% accuracy on MMRB2 and surpasses orders-of-magnitude larger models like Gemini-3-Pro. These results show that grounded verification and consistency-based training substantially improve multimodal judgment without increasing model scale.
title MJ1: Multimodal Judgment via Grounded Verification
topic Machine Learning
url https://arxiv.org/abs/2603.07990