Abstract: Test-time adaptation (TTA) aims to correct performance degradation of deep models under distribution shifts by updating models or inputs using unlabeled test data. Input-only diffusion-based TTA methods improve robustness to corruptions but rely on gradient guidance, limiting exploration and generalization across distortion types. We propose SteeringTTA, an inference-only framework that adapted Feynman-Kac steering to guide diffusion-based input adaptation with pseudo-label driven reward. SteeringTTA maintains multiple particle trajectories, steered by a combination of cumulative top-$K$ probabilities and an entropy schedule, to balance exploration and confidence. On ImageNet‐C, SteeringTTA consistently outperforms the baseline without any model updates or source data.