DeltaRubric: Generative Multimodal Reward Modeling via Joint Planning and Verification

Aligning Multimodal Large Language Models (MLLMs) requires reliable reward models, yet existing single-step evaluators can suffer from lazy judging, exploiting language priors over fine-grained visual verification. While rubric-based evaluation mitigates these biases in text-only settings, extending it to multimodal tasks is bottlenecked by the complexity of visual reasoning. The critical differences between responses often depend on instance-specific visual details. Robust evaluation requires dynamically synthesizing rubrics that isolate spatial and factual discrepancies. To address this, we introduce DeltaRubric, an approach that reformulates multimodal preference evaluation as a plan-and-execute process within a single MLLM. \dr operates in two steps: acting first as a $\text{Disagreement Planner}$, the model generates a neutral, instance-specific verification checklist. Transitioning into a $\text{Checklist Verifier}$, it executes these self-generated checks against the image and question to produce the final grounded judgment. We formulate \dr as a multi-role reinforcement learning problem, jointly optimizing planning and verification capabilities. Validated on Qwen3-VL 4B and 8B Instruct models, DeltaRubric achieves solid empirical gains, improving base-model overall accuracy on VL-RewardBench by $\text{+22.6}$ (4B) and $\text{+18.8}$ (8B) points and consistently surpassing no-rubric baselines. Beyond accuracy, DeltaRubric outperforms compute-matched inference scaling (verbose CoT and best-of-$N$), preference learning (DPO), and automated prompt optimization while generating fewer tokens, generalizes to a different backbone InternVL3 and to scalar rewards, and as a frozen reward model improves downstream policy training. The results demonstrate that decomposing evaluation into structured, verifiable steps leads to more reliable and generalizable multimodal reward modeling.