Online video recommendation systems often build binary labels based on play complete rate (i.e., the ratio of watch time to video duration), such as complete play and effective play, using them as implicit feedback for Click-Through Rate (CTR) prediction tasks to gauge user interest. Existing works tend to improve prediction accuracy by designing complex models, overlooking that a key cause of inaccurate predictions is the disorganization of instance representation space. To address this issue, we explore a novel approach using prototype learning to calibrate the instance representation space of deep recommendation models and propose a model-agnostic Contrastive Prototype Framework (CPF). Firstly, CPF partitions the instance space into different subspaces based on duration, then generates positive and negative prototype pairs for each subspace from pre-trained recommendation model. Subsequently, we map the instance representations to the prototype space and calibrate them by reducing the distance to the corresponding prototypes. Ultimately, the prediction is derived from the linear combination of the estimated values associated with each prototype. To prevent disorganization in the prototype space during training, we design contrastive and orthogonality losses to constrain the learning of prototypes. Additionally, we show that how CPF effectively addresses the duration bias from the perspective of causal intervention. Offline experiments on two datasets demonstrate that CPF improves recommendation accuracy over several baseline models in predicting five widely used implicit feedback labels. We have also deployed CPF on a short video platform, validating its effectiveness in real-world scenarios.

Growing mental health problems among college students underscore the urgent need to enhance the restorative effect of university campus. However, current evaluation tools lack systematic, scalable, and interpretable methods to quantify the psychological impact of open space design. This study proposes a multimodal perception system to predict the restorative effect of university open spaces by integrating visual, structural, and semantic features. We construct a large-scale dataset comprising 600 campus images and 12,147 subjective ratings collected through standardized psychological scales. Semantic segmentation is used to extract spatial visual indices from images, while semantic impressions are obtained via the Semantic Differential (SD) scale and converted into natural language descriptions. A dual-encoder alignment framework maps both index and text representations into a shared latent space, enabling cross-modal prediction of restorative scores. A Random Forest regressor is trained on this space to support score inference from either image- or text-based inputs. In addition, we apply a Rule-based Representation Learner (RRL) to extract interpretable spatial patterns associated with restorative outcomes. Experiments show that our method significantly outperforms traditional regression models, achieving an R2of 0.85 in predicting perceived restorative effect. The learned rules reveal both explicit visual drivers (e.g., greenery, sky openness) and implicit spatial logics (e.g., element interaction). This framework offers a lightweight and interpretable evaluation tool for health-oriented campus design, applicable across design and planning stages even without image data.