Interaction detection problem is to quantify the influence of any subset of input features that may potentially be an interaction.

Human-Object Interaction (HOI) detection is a fundamental task in computer vision, empowering machines to comprehend human-object relationships in diverse real-world scenarios. Recent advances in VLMs have significantly improved HOI detection by leveraging rich cross-modal representations. However, most existing VLM-based approaches rely heavily on additional training or prompt tuning, resulting in substantial computational overhead and limited scalability, particularly in long-tailed scenarios where rare interactions are severely underrepresented. In this paper, we propose the Adaptive Diversity Cache (ADC) module, a novel training-free and plug-and-play mechanism designed to mitigate long-tail bias in HOI detection. ADC constructs class-specific caches that accumulate high-confidence and diverse feature representations during inference. The method incorporates frequency-aware cache adaptation that favors rare categories and is designed to enable robust prediction calibration without requiring additional training or fine-tuning. Extensive experiments on HICO-DET and V-COCO datasets show that ADC consistently improves existing HOI detectors, achieving up to +8.57\% mAP gain on rare categories and +4.39\% on the full dataset, demonstrating its effectiveness in mitigating long-tail bias while preserving overall performance.

This document provides additional materials to supplement our main manuscript. The detection loss used for the output of human decoder ( i.e ., Moreover, an auxiliary loss is applied to the intermediate outputs of each decoder layer which contributes to improved results in the decoding process. We provide qualitative results of our method, including both success and failure cases in Fig. S2. Additionally, our model may be inefficient when it needs to deduce additional contextual cues.

For introduction to the image task, please see R1: Q2.

This document provides additional materials to supplement our main manuscript. The detection loss used for the output of human decoder ( i.e ., Moreover, an auxiliary loss is applied to the intermediate outputs of each decoder layer which contributes to improved results in the decoding process. We provide qualitative results of our method, including both success and failure cases in Fig. S2. Additionally, our model may be inefficient when it needs to deduce additional contextual cues.

Machine learning transparency calls for interpretable explanations of how inputs relate to predictions.