EvoRobust: Self-Improving MLLMs Robustness via Claim-Level Self-Verification

Multimodal large language models lose accuracy and make more unsupported statements when images are blurred, noisy, or heavily compressed. The robustness gap is hard to close by self-improvement, because the model being improved is also the only verifier and it reads exactly the same corrupted pixels. We introduce EvoRobust, which raises MLLM robustness via claim-level self-verification: it decomposes a free-form description into atomic claims, turns each claim into a targeted probe, re-answers every probe on the same corrupted image several times, and aggregates the resulting agreement into a claim-structured reward for Group Relative Policy Optimization. We are explicit about what this signal is: it estimates the agreement rate of a frozen verifier rather than a truth probability, repeated probing reduces decoding variance rather than recovering lost visual evidence, and policy optimization only needs the reward to rank candidates correctly. Trained on unlabeled degraded images, EvoRobust raises the R-Bench Overall score of Qwen3-VL-4B from $0.6671$ to $0.6782$, improves on averaged holistic self-scoring ($0.6715$) under an identical optimizer, and is first or tied-first in seven of nine corrupted settings across MMMB, MMStar, and RealWorldQA without reducing clean accuracy. We will release the code and configurations shortly.