Saved in:
Bibliographic Details
Main Authors: Kumar, Bhavesh, Feng, Dylan, Tang, Leonard
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.07990
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of 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.