Accurately estimating query-item relevance is vital for e-commerce ranking and conversion. While Large Language Models (LLMs) excel at reasoning, they often lack specialized knowledge required for long-tail or fast-evolving queries, necessitating Retrieval-Augmented Generation (RAG). However, production environments face three critical challenges: (1) external context is inherently noisy and inconsistent; (2) extreme latency budgets prohibit multi-stage processing or refinement; and (3) the model must simultaneously assess relevance and context-trust within a unified inference pass. We propose DyKnow-RAG, a reinforcement learning framework that teaches LLMs to learn to trust through dynamic utilization of external knowledge. Built on Group Relative Policy Optimization (GRPO), DyKnow-RAG utilizes a dual-group rollout strategy (parametric-only vs. with-context) and a posterior-driven inter-group advantage scaling mechanism. This enables the model to optimize context utilization without human process labels or extra inference overhead. Our pipeline further integrates structured Chain-of-Thought (CoT) and an uncertainty-prioritized RL pool to stabilize training. Offline evaluations show significant Macro-F1 and Accuracy gains, particularly on noise-sensitive query slices. Importantly, DyKnow-RAG has been deployed in Taobao's production system, serving hundreds of millions of active users and billions of daily search requests. Controlled A/B tests demonstrate consistent lifts in key business metrics, including GSB and Item Goodrate, while maintaining a p99 latency under 400ms. This work provides a scalable and deployable paradigm for operationalizing noisy RAG under extreme efficiency constraints of large-scale industrial search.

Query-product relevance prediction is fundamental to e-commerce search and has become even more critical in the era of AI-powered shopping, where semantic understanding and complex reasoning directly shape the user experience and business conversion. Large Language Models (LLMs) enable generative, reasoning-based approaches, typically aligned via supervised fine-tuning (SFT) or preference optimization methods like Direct Preference Optimization (DPO). However, the increasing complexity of business rules and user queries exposes the inability of existing methods to endow models with robust reasoning capacity for long-tail and challenging cases. Efforts to address this via reinforcement learning strategies like Group Relative Policy Optimization (GRPO) often suffer from sparse terminal rewards, offering insufficient guidance for multi-step reasoning and slowing convergence. To address these challenges, we propose TaoSR-AGRL, an Adaptive Guided Reinforcement Learning framework for LLM-based relevance prediction in Taobao Search Relevance. TaoSR-AGRL introduces two key innovations: (1) Rule-aware Reward Shaping, which decomposes the final relevance judgment into dense, structured rewards aligned with domain-specific relevance criteria; and (2) Adaptive Guided Replay, which identifies low-accuracy rollouts during training and injects targeted ground-truth guidance to steer the policy away from stagnant, rule-violating reasoning patterns toward compliant trajectories. TaoSR-AGRL was evaluated on large-scale real-world datasets and through online side-by-side human evaluations on Taobao Search. It consistently outperforms DPO and standard GRPO baselines in offline experiments, improving relevance accuracy, rule adherence, and training stability. The model trained with TaoSR-AGRL has been successfully deployed in the main search scenario on Taobao, serving hundreds of millions of users.

Text relevance or text matching of query and product is an essential technique for the e-commerce search system to ensure that the displayed products can match the intent of the query. Many studies focus on improving the performance of the relevance model in search system. Recently, pre-trained language models like BERT have achieved promising performance on the text relevance task. While these models perform well on the offline test dataset, there are still obstacles to deploy the pre-trained language model to the online system as their high latency. The two-tower model is extensively employed in industrial scenarios, owing to its ability to harmonize performance with computational efficiency. Regrettably, such models present an opaque ''black box'' nature, which prevents developers from making special optimizations. In this paper, we raise deep Bag-o f-Words (DeepBoW) model, an efficient and interpretable relevance architecture for Chinese e-commerce. Our approach proposes to encode the query and the product into the sparse BoW representation, which is a set of word-weight pairs. The weight means the important or the relevant score between the corresponding word and the raw text. The relevance score is measured by the accumulation of the matched word between the sparse BoW representation of the query and the product. Compared to popular dense distributed representation that usually suffers from the drawback of black-box, the most advantage of the proposed representation model is highly explainable and interventionable, which is a superior advantage to the deployment and operation of online search engines. Moreover, the online efficiency of the proposed model is even better than the most efficient inner product form of dense representation. The proposed model is experimented on three different datasets for learning the sparse BoW representations, including the human-annotation set, the search-log set and the click-through set. Then the models are evaluated by experienced human annotators. Both the auto metrics and the online evaluations show our DeepBoW model achieves competitive performance while the online inference is much more efficient than the other models. Our DeepBoW model has already deployed to the biggest Chinese e-commerce search engine Taobao and served the entire search traffic for over 6 months.

To enhance the effectiveness of matching user requests with millions of online products, practitioners invest significant efforts in developing semantic relevance models on large-scale e-commerce platforms. Generally, such semantic relevance models are formulated as text-matching approaches, which measure the relevance between users' search queries and the titles of candidate items (i.e., products). However, we argue that conventional relevance methods may lead to sub-optimal performance due to the limited information provided by the titles of candidate items. To alleviate this issue, we suggest incorporating additional information about candidate items from multiple aspects, including their attributes and images. This could supplement the information that may not be fully provided by titles alone. To this end, we propose a multi-aspect semantic relevance model that takes into account the match between search queries and the title, attribute and image information of items simultaneously. The model is further enhanced through pre-training using several well-designed self-supervised and weakly-supervised tasks. Furthermore, the proposed model is fine-tuned using annotated data and distilled into a representation-based architecture for efficient online deployment. Experimental results show the proposed approach significantly improves relevance and leads to considerable enhancements in business metrics.

Text relevance or text matching of query and product is an essential technique for e-commerce search engine, which helps users find the desirable products and is also crucial to ensuring user experience. A major difficulty for e-commerce text relevance is the severe vocabulary gap between query and product. Recently, neural networks have been the mainstream for the text matching task owing to the better performance for semantic matching. Practical e-commerce relevance models are usually representation-based architecture, which can pre-compute representations offline and are therefore online efficient. Interaction-based models, although can achieve better performance, are mostly time-consuming and hard to be deployed online. Recently BERT has achieved significant progress on many NLP tasks including text matching, and it is of great value but also big challenge to deploy BERT to the e-commerce relevance task. To realize this goal, we propose ReprBERT, which has the advantages of both excellent performance and low latency, by distilling the interaction-based BERT model to a representation-based architecture. To reduce the performance decline, we investigate the key reasons and propose two novel interaction strategies to resolve the absence of representation interaction and low-level semantic interaction. Finally, ReprBERT can achieve only about 1.5% AUC loss from the interaction-based BERT, but has more than 10% AUC improvement compared to previous state-of-the-art representation-based models. ReprBERT has already been deployed on the search engine of Taobao and serving the entire search traffic, achieving significant gain of user experience and business profit.

The search engine plays a fundamental role in online e-commerce systems, to help users find the products they want from the massive product collections. Relevance is an essential requirement for e-commerce search, since showing products that do not match search query intent will degrade user experience. With the existence of vocabulary gap between user language of queries and seller language of products, measuring semantic relevance is necessary and neural networks are engaged to address this task. However, semantic relevance is different from click-through rate prediction in that no direct training signal is available. Most previous attempts learn relevance models from user click-through data that are cheap and abundant. Unfortunately, click behavior is noisy and misleading, which is affected by not only relevance but also factors including price, image and attractive titles. Therefore, it is challenging but valuable to learn relevance models from click-through data. In this paper, we propose a new relevance learning framework that concentrates on how to train a relevance model from the weak supervision of click-through data. Different from previous efforts that treat samples as either relevant or irrelevant, we construct more fine-grained samples for training. We propose a novel way to consider samples of different relevance confidence, and come up with a new training objective to learn a robust relevance model with desirable score distribution. The proposed model is evaluated on offline annotated data and online A/B testing, and it achieves both promising performance and high computational efficiency. The model has already been deployed online, serving the search traffic of Taobao for over a year.

Multimodal Machine Translation (MMT) aims to introduce information from other modality, generally static images, to improve the translation quality.Previous works propose various incorporation methods, but most of them do not consider the relative importance of multiple modalities.In MMT, equally treating text and images may encode too much irrelevant information from images which may introduce noise.In this paper, we propose the multimodal self-attention in Transformer to solve the issues above.The proposed method learns the representations of images based on the text, which avoids encoding irrelevant information in images.Experiments and visualization analysis demonstrate that our model benefits from visual information and substantially outperforms previous works and competitive baselines in terms of various metrics.

The graph-to-sequence (Graph2Seq) learning aims to transduce graph-structured representations to word sequences for text generation.Recent studies propose various models to encode graph structure.However, most previous works ignore the indirect relations between distance nodes, or treat indirect relations and direct relations in the same way.In this paper, we propose the Heterogeneous Graph Transformer to independently model the different relations in the individual subgraphs of the original graph, including direct relations, indirect relations and multiple possible relations between nodes.Experimental results show that our model strongly outperforms the state of the art on all four standard benchmarks of AMRto-text generation and syntax-based neural machine translation.

AMR-to-text generation is a challenging task of generating texts from graph-based semantic representations. Recent studies formalize this task a graph-to-sequence learning problem and use various graph neural networks to model graph structure. In this paper, we propose a novel approach that generates texts from AMR graphs while reconstructing the input graph structures. Our model employs graph attention mechanism to aggregate information for encoding the inputs. Moreover, better node representations are learned by optimizing two simple but effective auxiliary reconstruction objectives: link prediction objective which requires predicting the semantic relationship between nodes, and distance prediction objective which requires predicting the distance between nodes. Experimental results on two benchmark datasets show that our proposed model improves considerably over strong baselines and achieves new state-of-the-art.

The synchronization of output torque is an important issue during human-robot interaction. Since the human-robot synchronization may be affected by the delay between the human's voluntary torque and the robot's assistive torque when a torque sensor is applied to detect human intention, an electromyography (EMG)-based admittance controller (EAC) was proposed to improve human-robot synchronization compared with that achieved by the use of a torque-sensing-based admittance controller (TAC). Simulations and experiments were conducted to investigate the performance of the EAC and the TAC. The simulation results indicated that the delay between the human's voluntary torque and the exoskeleton robot's assistive torque significantly degraded the performance of the human-robot cooperation movement when the TAC was used. The experimental results indicated that the jerk value, interaction torque, and EMG level of the tibialis anterior obtained with the EAC were significantly smaller than those with the TAC. Compared with the TAC, the EAC has an advantage in improving human-robot cooperation movement because the EAC can reduce the delay between the human's voluntary torque and the exoskeleton robot's assistive torque.