Hydra: An Agentic Reasoning Approach for Enhancing Adversarial Robustness and Mitigating Hallucinations in Vision-Language Models

Hydra is an agentic reasoning approach that enhances adversarial robustness and mitigates hallucinations in Vision-Language Models through multi-agent deliberation and structured reasoning.