Ranking relevance is a fundamental task in search engines, aiming to identify the items most relevant to a given user query. Traditional relevance models typically produce scalar scores or directly predict relevance labels, limiting both interpretability and the modeling of complex relevance signals. Inspired by recent advances in Chain-of-Thought (CoT) reasoning for complex tasks, we investigate whether explicit reasoning can enhance both interpretability and performance in relevance modeling. However, existing reasoning-based Generative Relevance Models (GRMs) primarily rely on supervised fine-tuning on large amounts of human-annotated or synthetic CoT data, which often leads to limited generalization. Moreover, domain-agnostic, free-form reasoning tends to be overly generic and insufficiently grounded, limiting its potential to handle the diverse and ambiguous cases prevalent in open-domain search. In this work, we formulate relevance modeling in Xiaohongshu search as a reasoning task and introduce a Reinforcement Learning (RL)-based training framework to enhance the grounded reasoning capabilities of GRMs. Specifically, we incorporate practical business-specific relevance criteria into the multi-step reasoning prompt design and propose Stepwise Advantage Masking (SAM), a lightweight process-supervision strategy which facilitates effective learning of these criteria through improved credit assignment. To enable industrial deployment, we further distill the large-scale RL-tuned model to a lightweight version suitable for real-world search systems. Extensive offline evaluations and online A/B tests demonstrate that our approach consistently delivers significant improvements across key relevance and business metrics, validating its effectiveness, robustness, and practicality for large-scale industrial search systems.

Benefiting from the strong reasoning capabilities, Large language models (LLMs) have demonstrated remarkable performance in recommender systems. Various efforts have been made to distill knowledge from LLMs to enhance collaborative models, employing techniques like contrastive learning for representation alignment. In this work, we prove that directly aligning the representations of LLMs and collaborative models is suboptimal for enhancing downstream recommendation tasks performance, based on the information theorem. Consequently, the challenge of effectively aligning semantic representations between collaborative models and LLMs remains unresolved. Inspired by this viewpoint, we propose a novel plug-and-play alignment framework for LLMs and collaborative models. Specifically, we first disentangle the latent representations of both LLMs and collaborative models into specific and shared components via projection layers and representation regularization. Subsequently, we perform both global and local structure alignment on the shared representations to facilitate knowledge transfer. Additionally, we theoretically prove that the specific and shared representations contain more pertinent and less irrelevant information, which can enhance the effectiveness of downstream recommendation tasks. Extensive experimental results on benchmark datasets demonstrate that our method is superior to existing state-of-the-art algorithms.

Recent neural models for video captioning usually employed an encoder-decoder framework. However, most approaches either neglected the spatial and temporal interactions between objects in a video or implicitly modelled the interactions, resulting in less desired performance. In this paper, we propose a novel relation-aware graph learning framework. It explicitly models both spatial and temporal relations for objects. In particular, a relation-aware graph is designed to depict the spatial relations between different objects in a scene. Parallelly, a temporal graph network is designed to perform relational reasoning for the same objects in adjacent frames. Features of both types of relations are learned and fused for the follow-up language decoder. Experiments on two bench-mark datasets show the effectiveness of our framework. It achieves state-of-the-art performance with CIDEr scores on MSVD and MSR-VTT.

Visual place recognition is to estimate the geographical location of a given image, which is usually addressed by recognizing its similar reference images from a database. The reference images are usually retrieved via similarity search using global descriptor, and the local descriptors are used to re-rank the initial retrieved candidates. The local descriptors re-ranking can significantly improve the accuracy of global retrieval but comes at a high computational cost. To achieve a good trade-off between accuracy and efficiency, we propose an Efficient Transformer for Re-ranking (ETR), utilizing both global and local descriptors to re-rank the top candidates in a single shot. In contrast to traditional re-ranking methods, we leverage self-attention to capture relationships between local descriptors in a single image and cross-attention to explore the similarity of the image pairs. We show that the proposed model can be regarded as a general re-ranking algorithm for significantly boosting the performance of other global-only retrieval methods. Extensive experimental results show that our method outperforms state-of-the-arts and is orders of magnitude faster in terms of computational efficiency.

Video captioning aims at generating a natural language de-scription for a given video clip including not only salient sce-narios but also contextual scenarios. The former reveal the highlight of a video and are usually the focus of most existing captioning methods. The latter, however, are not well ex-plored and even ignored easily, though they may provide cer-tain detailed and latent information that can help with a better understanding of the video. To effectively exploit the infor-mation contained in both, a novel video captioning network is proposed. It has two key modules: Cross-Modality Selection (CMS) and Saliency-Context Adaptive Decoder (SCAD). Specifically, CMS mainly focuses on utilizing the semantic information to distinguish saliency and context. Meanwhile, SCAD adaptively identifies both the saliency and context to generate more detailed and precise captions. Experiments on two benchmark datasets, i.e., MSVD and MSR-VTT, demon-strate the effectiveness of our model through the comparison with state-of-the-art methods.