GRPO vs PPO in LLaVA-UHD: Robustness to Adversarial Perturbations on SEED-Bench-R1

This report synthesises findings from 1 peer-reviewed paper addressing the following research question: What is the impact of replacing PPO with GRPO on the multimodal reasoning robustness of LLaVA-UHD models when evaluated against adversarial perturbations in SEED-Bench-R1. Multimodal Large Language Models (MLLMs) have achieved remarkable progress in various vision-language tasks, including visual recognition, visual grounding, and video understanding. However, as their capabilities primarily derive from large-scale pretraining with limited. 6 claims were extracted from source literature; 6 were independently verified against retrieved documents. An automated multi-reviewer quality assessment produced a score of 8.2/10. This report is a machine-generated literature synthesis and does not constitute original research.

Research goal: What is the impact of replacing PPO with GRPO on the multimodal reasoning robustness of LLaVA-UHD models when evaluated against adversarial perturbations in SEED-Bench-R1?

Autonomous literature synthesis. Automated review score: 8.2/10. Full text and citation available at Assignee Research.