Robust-R1: Degradation-Aware Reasoning for Robust Visual Understanding

TL;DR

Robust-R1 tackles the fragility of multimodal LLMs under real-world visual degradations by introducing degradation-aware reasoning built on structured reasoning chains. The method combines supervised fine-tuning, reward-driven degradation-parameter alignment, and adaptive reasoning-length scaling to maintain robust outputs. A dedicated 11K degradation dataset across four processing stages supports training and evaluation. Empirical results on R-Bench and adversarial degradation benchmarks show state-of-the-art robustness and improved efficiency compared with existing general and robust MLLMs.

Abstract

Multimodal Large Language Models struggle to maintain reliable performance under extreme real-world visual degradations, which impede their practical robustness. Existing robust MLLMs predominantly rely on implicit training/adaptation that focuses solely on visual encoder generalization, suffering from limited interpretability and isolated optimization. To overcome these limitations, we propose Robust-R1, a novel framework that explicitly models visual degradations through structured reasoning chains. Our approach integrates: (i) supervised fine-tuning for degradation-aware reasoning foundations, (ii) reward-driven alignment for accurately perceiving degradation parameters, and (iii) dynamic reasoning depth scaling adapted to degradation intensity. To facilitate this approach, we introduce a specialized 11K dataset featuring realistic degradations synthesized across four critical real-world visual processing stages, each annotated with structured chains connecting degradation parameters, perceptual influence, pristine semantic reasoning chain, and conclusion. Comprehensive evaluations demonstrate state-of-the-art robustness: Robust-R1 outperforms all general and robust baselines on the real-world degradation benchmark R-Bench, while maintaining superior anti-degradation performance under multi-intensity adversarial degradations on MMMB, MMStar, and RealWorldQA.

Figures (6)

• Figure 1: Comparison with other existing robustness enhancement approaches. (A) is based on implicit training/adaptation, which only considers the visual encoder feature alignment. (B) is ours, and we explicitly integrate the degradation-aware reasoning chain into MLLM.
• Figure 2: Overview of Robust-R1. (A) Supervised Fine-Tuning (SFT): we train the model using reasoning data to equip it with basic degradation-aware reasoning capability; (B) Reinforcement Learning (RL): we propose two reward functions to (i) align precise degradation-aware space while (ii) adaptively scaling to suitable reasoning lengths based on degradation intensity.
• Figure 3: Correlation between degradation intensity and reasoning chain length on Seed-1.5-VL guo2025seed1. Higher degradation intensities require longer chains to maintain accuracy, even multi-step reasoning.
• Figure 4: Data generation pipeline. The original images undergo various real-world processing stages, where multiple degradations are randomly added to obtain degraded images and their corresponding degradation <TYPE>s. Based on these and the original question-answering pairs (QAs), the pipeline progressively generates <INFLUENCE>, <REASONING>, and <CONCLUSION>. Finally, the reasoning chain is scaling according to different intensities to achieve optimal efficiency.
• Figure 5: Qualitative evaluation for anti-degradation. Ours (SFT and RL) can provide robust and efficient result.