We present VisualTreeSearch, a fully-deployed system for visualizing and understanding web agent test-time scaling. While test-time search algorithms substantially improve web agent success rates, they remain confined to research contexts with limited practical deployment. Our system bridges this gap with three key contributions: (1) a production-ready solution with cloud-based architecture, (2) an efficient API-based state reset mechanism that reduces state reset time from 50 to 2 s, and (3) an interactive web UI that transparently demonstrates the agent's decision-making process. VisualTreeSearch provides an intuitive framework for both researchers and users to understand tree search execution in web agents.

The development of digital twins for geotechnical structures necessitates the real-time updates of three-dimensional (3D) virtual models (e.g. numerical finite element method (FEM) model) to accurately predict time-varying geotechnical responses (e.g. consolidation settlement) in a 3D spatial domain. However, traditional 3D numerical model updating approaches are computationally prohibitive and therefore difficult to update the 3D responses in real time. To address these challenges, this study proposes a novel machine learning framework called sparse dictionary learning (T-3D-SDL) for real-time updating of time-varying 3D geotechnical responses. In T-3D-SDL, a concerned dataset (e.g. time-varying 3D settlement) is approximated as a linear superposition of dictionary atoms generated from 3D random FEM analyses. Field monitoring data are then used to identify non-trivial atoms and estimate their weights within a Bayesian framework for model updating and prediction. The proposed approach enables the real-time update of temporally varying settlements with a high 3D spatial resolution and quantified uncertainty as field monitoring data evolve. The proposed approach is illustrated using an embankment construction project. The results show that the proposed approach effectively improves settlement predictions along temporal and 3D spatial dimensions, with minimal latency (e.g. within minutes), as monitoring data appear. In addition, the proposed approach requires only a reasonably small number of 3D FEM model evaluations, avoids the use of widely adopted yet often criticized surrogate models, and effectively addresses the limitations (e.g. computational inefficiency) of existing 3D model updating approaches.

Over the past two years, GAI has evolved rapidly, influencing various fields including social and e-commerce Recsys. Despite exciting advances, landing these innovations in real-world Recsys remains challenging due to the sophistication of modern industrial product and systems. Our tutorial begins with a brief overview of building industrial Recsys and GAI fundamentals, followed by the ongoing efforts and opportunities to enhance personalized recommendations with foundation models. We then explore the integration of curation capabilities into Recsys, such as repurposing raw content, incorporating external knowledge, and generating personalized insights/explanations to foster transparency and trust. Next, the tutorial illustrates how AI agents can transform Recsys through interactive reasoning and action loops, shifting away from traditional passive feedback models. Finally, we shed insights on real-world solutions for human-AI alignment and responsible GAI practices. A critical component of the tutorial is detailing the AI, Infrastructure, LLMOps, and Product roadmap (including the evaluation and responsible AI practices) derived from the production solutions in LinkedIn, Amazon, TikTok, and Microsoft. While GAI in Recsys is still in its early stages, this tutorial provides valuable insights and practical solutions for the Recsys and GAI communities.

We introduce LiteWebAgent, an open-source suite for VLM-based web agent applications. LiteWebAgent addresses a critical gap in the web agent ecosystem by providing an extensible core agent framework featuring planning, memory, and tree search capabilities, alongside a production-ready solution that combines minimal serverless backend configuration and intuitive user and browser interfaces. For the core LiteWebAgent agent framework, we implemented a simple yet effective baseline using recursive function calling, providing decoupled action generation and action grounding. In addition, we integrate advanced research components such as agent planning, agent workflow memory, and tree search in a modular and extensible manner. We then integrate the LiteWebAgent agent framework with frontend and backend systems in two deployment formats: (1) a production Vercelbased web application that provides users with an agent-controlled remote browser, and (2) a Chrome extension that leverages LiteWebAgent's API to control an existing Chrome browser via CDP (Chrome DevTools Protocol). The core LiteWebAgent framework is available at https://github.com/PathOnAI/ LiteWebAgent, with deployed frontend at https://lite-web- agent.vercel.app/.

Recently, generative AI (GAI), with their emerging capabilities, have presented unique opportunities for augmenting and revolutionizing industrial recommender systems (Recsys). Despite growing research efforts at the intersection of these fields, the integration of GAI into industrial Recsys remains in its infancy, largely due to the intricate nature of modern industrial Recsys infrastructure, operations, and product sophistication. Drawing upon our experiences in successfully integrating GAI into several major social and e-commerce platforms, this survey aims to comprehensively examine the underlying system and AI foundations, solution frameworks, connections to key research advancements, as well as summarize the practical insights and challenges encountered in the endeavor to integrate GAI into industrial Recsys. As pioneering work in this domain, we hope outline the representative developments of relevant fields, shed lights on practical GAI adoptions in the industry, and motivate future research.

Aiming at a better understanding of the search goals in the user search sessions, recent query recommender systems explicitly model the reformulations of queries, which hopes to estimate the intents behind these reformulations and thus benefit the next-query recommendation. However, in real-world e-commercial search scenarios, user intents are much more complicated and may evolve dynamically. Existing methods merely consider trivial reformulation intents from semantic aspects and fail to model dynamic reformulation intent flows in search sessions, leading to sub-optimal capacities to recommend desired queries. To deal with these limitations, we first explicitly define six types of query reformulation intents according to the desired products of two consecutive queries. We then apply two self-attentive encoders on top of two pre-trained large language models to learn the transition dynamics from semantic query and intent reformulation sequences, respectively. We develop an intent-aware query decoder to utilize the predicted intents for suggesting the next queries. We instantiate such a framework with an Intent-aware Variational AutoEncoder (IVAE) under deployment at Amazon. We conduct comprehensive experiments on two real-world e-commercial datasets from Amazon and one public dataset from BestBuy. Specifically, IVAE improves the Recall@15 by 25.44% and 60.47% on two Amazon datasets and 13.91% on BestBuy, respectively.

Recent research showed promising results on combining pretrained language models (LMs) with canonical utterance for few-shot semantic parsing. The canonical utterance is often lengthy and complex due to the compositional structure of formal languages. Learning to generate such canonical utterance requires significant amount of data to reach high performance. Fine-tuning with only few-shot samples, the LMs can easily forget pretrained knowledge, overfit spurious biases, and suffer from compositionally out-of-distribution generalization errors. To tackle these issues, we propose a novel few-shot semantic parsing method -- SeqZero. SeqZero decomposes the problem into a sequence of sub-problems, which correspond to the sub-clauses of the formal language. Based on the decomposition, the LMs only need to generate short answers using prompts for predicting sub-clauses. Thus, SeqZero avoids generating a long canonical utterance at once. Moreover, SeqZero employs not only a few-shot model but also a zero-shot model to alleviate the overfitting. In particular, SeqZero brings out the merits from both models via ensemble equipped with our proposed constrained rescaling. SeqZero achieves SOTA performance of BART-based models on GeoQuery and EcommerceQuery, which are two few-shot datasets with compositional data split.

E-commerce query understanding is the process of inferring the shopping intent of customers by extracting semantic meaning from their search queries. The recent progress of pre-trained masked language models (MLM) in natural language processing is extremely attractive for developing effective query understanding models. Specifically, MLM learns contextual text embedding via recovering the masked tokens in the sentences. Such a pre-training process relies on the sufficient contextual information. It is, however, less effective for search queries, which are usually short text. When applying masking to short search queries, most contextual information is lost and the intent of the search queries may be changed. To mitigate the above issues for MLM pre-training on search queries, we propose a novel pre-training task specifically designed for short text, called Extended Token Classification (ETC). Instead of masking the input text, our approach extends the input by inserting tokens via a generator network, and trains a discriminator to identify which tokens are inserted in the extended input. We conduct experiments in an E-commerce store to demonstrate the effectiveness of ETC.

Graph data augmentation has been used to improve generalizability of graph machine learning. However, by only applying fixed augmentation operations on entire graphs, existing methods overlook the unique characteristics of communities which naturally exist in the graphs. For example, different communities can have various degree distributions and homophily ratios. Ignoring such discrepancy with unified augmentation strategies on the entire graph could lead to sub-optimal performance for graph data augmentation methods. In this paper, we study a novel problem of automated graph data augmentation for node classification from the localized perspective of communities. We formulate it as a bilevel optimization problem: finding a set of augmentation strategies for each community, which maximizes the performance of graph neural networks on node classification. As the bilevel optimization is hard to solve directly and the search space for community-customized augmentations strategy is huge, we propose a reinforcement learning framework AutoGDA that learns the local-optimal augmentation strategy for each community sequentially. Our proposed approach outperforms established and popular baselines on public node classification benchmarks as well as real industry e-commerce networks by up to +12.5% accuracy.

As training deep learning models on large dataset takes a lot of time and resources, it is desired to construct a small synthetic dataset with which we can train deep learning models sufficiently. There are recent works that have explored solutions on condensing image datasets through complex bi-level optimization. For instance, dataset condensation (DC) matches network gradients w.r.t. large-real data and small-synthetic data, where the network weights are optimized for multiple steps at each outer iteration. However, existing approaches have their inherent limitations: (1) they are not directly applicable to graphs where the data is discrete; and (2) the condensation process is computationally expensive due to the involved nested optimization. To bridge the gap, we investigate efficient dataset condensation tailored for graph datasets where we model the discrete graph structure as a probabilistic model. We further propose a one-step gradient matching scheme, which performs gradient matching for only one single step without training the network weights. Our theoretical analysis shows this strategy can generate synthetic graphs that lead to lower classification loss on real graphs. Extensive experiments on various graph datasets demonstrate the effectiveness and efficiency of the proposed method. In particular, we are able to reduce the dataset size by 90% while approximating up to 98% of the original performance and our method is significantly faster than multi-step gradient matching (e.g. $15$× in CIFAR10 for synthesizing 500 graphs).

With the increasing demands on e-commerce platforms, numerous user action history is emerging. Those enriched action records are vital to understand users' interests and intents. Recently, prior works for user behavior prediction mainly focus on the interactions with product-side information. However, the interactions with search queries, which usually act as a bridge between users and products, are still under investigated. In this paper, we explore a new problem named temporal event forecasting, a generalized user behavior prediction task in a unified query product evolutionary graph, to embrace both query and product recommendation in a temporal manner. To fulfill this setting, there involves two challenges: (1) the action data for most users is scarce; (2) user preferences are dynamically evolving and shifting over time. To tackle those issues, we propose a novel Retrieval-Enhanced Temporal Event (RETE) forecasting framework. Unlike existing methods that enhance user representations via roughly absorbing information from connected entities in the whole graph, RETE efficiently and dynamically retrieves relevant entities centrally on each user as high-quality subgraphs, preventing the noise propagation from the densely evolutionary graph structures that incorporate abundant search queries. And meanwhile, RETE autoregressively accumulates retrieval-enhanced user representations from each time step, to capture evolutionary patterns for joint query and product prediction. Empirically, extensive experiments on both the public benchmark and four real-world industrial datasets demonstrate the effectiveness of the proposed RETE method.

Weak supervision has shown promising results in many natural language processing tasks, such as Named Entity Recognition (NER). Existing work mainly focuses on learning deep NER models only with weak supervision, i.e., without any human annotation, and shows that by merely using weakly labeled data, one can achieve good performance, though still underperforms fully supervised NER with manually/strongly labeled data. In this paper, we consider a more practical scenario, where we have both a small amount of strongly labeled data and a large amount of weakly labeled data. Unfortunately, we observe that weakly labeled data does not necessarily improve, or even deteriorate the model performance (due to the extensive noise in the weak labels) when we train deep NER models over a simple or weighted combination of the strongly labeled and weakly labeled data. To address this issue, we propose a new multi-stage computational framework - NEEDLE with three essential ingredients: (1) weak label completion, (2) noise-aware loss function, and (3) final finetuning over the strongly labeled data. Through experiments on E-commerce query NER and Biomedical NER, we demonstrate that NEEDLE can effectively suppress the noise of the weak labels and outperforms existing methods. In particular, we achieve new SOTA F1-scores on 3 Biomedical NER datasets: BC5CDRchem 93.74, BC5CDR-disease 90.69, NCBI-disease 92.28.

We study the problem of query attribute value extraction, which aims to identify named entities from user queries as diverse surface form attribute values and afterward transform them into formally canonical forms. Such a problem consists of two phases: named entity recognition (NER) and attribute value normalization (AVN). However, existing works only focus on the NER phase but neglect equally important AVN. To bridge this gap, this paper proposes a unified query attribute value extraction system in e-commerce search named QUEACO, which involves both two phases. Moreover, by leveraging large-scale weakly-labeled behavior data, we further improve the extraction performance with less supervision cost. Specifically, for the NER phase, QUEACO adopts a novel teacher-student network, where a teacher network that is trained on the strongly-labeled data generates pseudo-labels to refine the weakly-labeled data for training a student network. Meanwhile, the teacher network can be dynamically adapted by the feedback of the student's performance on strongly-labeled data to maximally denoise the noisy supervisions from the weak labels. For the AVN phase, we also leverage the weakly-labeled query-to-attribute behavior data to normalize surface form attribute values from queries into canonical forms from products. Extensive experiments on a real-world large-scale E-commerce dataset demonstrate the effectiveness of QUEACO.

Sequence labeling aims to predict a fine-grained sequence of labels for the text. However, such formulation hinders the effectiveness of supervised methods due to the lack of token-level annotated data. This is exacerbated when we meet a diverse range of languages. In this work, we explore multilingual sequence labeling with minimal supervision using a single unified model for multiple languages. Specifically, we propose a Meta Teacher-Student (MetaTS) Network, a novel meta learning method to alleviate data scarcity by leveraging large multilingual unlabeled data. Prior teacher-student frameworks of self-training rely on rigid teaching strategies, which may hardly produce high-quality pseudo-labels for consecutive and interdependent tokens. On the contrary, MetaTS allows the teacher to dynamically adapt its pseudo-annotation strategies by the student's feedback on the generated pseudo-labeled data of each language and thus mitigate error propagation from noisy pseudo-labels. Extensive experiments on both public and real-world multilingual sequence labeling datasets empirically demonstrate the effectiveness of MetaTS(1).

Exploiting label hierarchies has become a promising approach to tackling the zero-shot multi-label text classification (ZS-MTC) problem. Conventional methods aim to learn a matching model between text and labels, using a graph encoder to incorporate label hierarchies to obtain effective label representations (Rios and Kavuluru, 2018). More recently, pretrained models like BERT (Devlin et al., 2018) have been used to convert classification tasks into a textual entailment task (Yin et al., 2019). This approach is naturally suitable for the ZS-MTC task. However, pretrained models are underexplored in the existing work because they do not generate individual vector representations for text or labels, making it unintuitive to combine them with conventional graph encoding methods. In this paper, we explore to improve pretrained models with label hierarchies on the ZS-MTC task. We propose a Reinforced Label Hierarchy Reasoning (RLHR) approach to encourage interdependence among labels in the hierarchies during training. Meanwhile, to overcome the weakness of flat predictions, we design a rollback algorithm that can remove logical errors from predictions during inference. Experimental results on three real-life datasets show that our approach achieves better performance and outperforms previous non-pretrained methods on the ZS-MTC task.

In this paper, we focus on data augmentation for the extreme multi-label classification (XMC) problem. One of the most challenging issues of XMC is the long tail label distribution where even strong models suffer from insufficient supervision. To mitigate such label bias, we propose a simple and effective augmentation framework and a new state-of-the-art classifier. Our augmentation framework takes advantage of the pre-trained GPT-2 model to generate label-invariant perturbations of the input texts to augment the existing training data. As a result, it present substantial improvements over baseline models. Our contributions are two-factored: (1) we introduce a new state-of-the-art classifier that uses label attention with RoBERTa and combine it with our augmentation framework for further improvement; (2) we present a broad study on how effective are different augmentation methods in the XMC task.

We model the impact of weather and arrival demand on ground delay program (GDP) incidence. We use Support Vector Machine (SVM) to analyze how regional convective weather affects GDP incidence and find the impact depends on both distance and direction of convective activity from the airport. We then train and compare the performance of logistic regression (LR) and random forest (RF) in predicting GDP incidence using an SVM-generated regional weather variable, local weather and arrival demand. Generally, RF outperforms LR. Convective weather is the most important factor in predicting GDP incidence at Atlanta International Airport (ATL), while arrival demand has greater impact for the other airports studied. We also examined model transferability across different airports. Lastly, we build GDP duration prediction models to enable a user to assess how long a GDP is likely to continue, if it is in effect in a given hour.

Agent-based modeling in transportation problems requires detailed information on each of the agents that represent the population in the region of a study. To extend the agent-based transportation modeling with social influence, a connected synthetic population with both synthetic features and its social networks need to be simulated. However, either the traditional manually-collected household survey data (ACS) or the recent large-scale passively-collected Call Detail Records (CDR) alone lacks features. This work proposes an algorithmic procedure that makes use of both traditional survey data as well as digital records of networking and human behavior to generate connected synthetic populations. The generated populations coupled with recent advances in graph (social networks) algorithms can be used for testing transportation simulation scenarios with different social factors.

Given the linked account graph with tens of millions of vertices, and a list of confirmed risky accounts, how can we quickly find a short list of potentially risky accounts for further human expert investigation? Most mainstream graph-based fraud detection algorithms focusing on detecting dense blocks of fake follows, or fake reviews from the social media graph, however, do not align well with the answer to the question. Here we hypothesize that fraud accounts share dense connections within a "fraud community", but have less so with accounts outside of the community. We propose GraphRAD, a risky account detection system based on local graph clustering algorithms. Our experiments show that from a real-world account graph of ≈ 60 million vertices and ≈ 500 million edges, GraphRAD was able to catch 67 previously unidentified fraud accounts by proposing 28 small-scale local communities, which is significantly more effective than the baseline model.