Why we think this paper is great for you:
This paper directly addresses pre-ranking in industrial recommendation systems, offering valuable insights into optimizing the efficiency of your travel ranking and recommendation processes. It's highly relevant to improving the performance of your travel platforms.

Abstract
In industrial recommendation systems, pre-ranking models based on deep neural networks (DNNs) commonly adopt a sequential execution framework: feature fetching and model forward computation are triggered only after receiving candidates from the upstream retrieval stage. This design introduces inherent bottlenecks, including redundant computations of identical users/items and increased latency due to strictly sequential operations, which jointly constrain the model's capacity and system efficiency. To address these limitations, we propose the Asynchronous Inference Framework (AIF), a cost-effective computational architecture that decouples interaction-independent components, those operating within a single user or item, from real-time prediction. AIF reorganizes the model inference process by performing user-side computations in parallel with the retrieval stage and conducting item-side computations in a nearline manner. This means that interaction-independent components are calculated just once and completed before the real-time prediction phase of the pre-ranking stage. As a result, AIF enhances computational efficiency and reduces latency, freeing up resources to significantly improve the feature set and model architecture of interaction-independent components. Moreover, we delve into model design within the AIF framework, employing approximated methods for interaction-dependent components in online real-time predictions. By co-designing both the framework and the model, our solution achieves notable performance gains without significantly increasing computational and latency costs. This has enabled the successful deployment of AIF in the Taobao display advertising system.

AI Summary

• AIF decouples interaction-independent computations (user/item-side) from real-time prediction, enabling asynchronous pre-computation to reduce latency and redundant computation in pre-ranking. [3]
• The co-design of the AIF framework and model architecture allows for richer feature sets and more advanced models in pre-ranking, achieving significant performance gains (e.g., +8.72% CTR, +5.80% RPM) within strict resource constraints. [3]
• Asynchronous Inference Framework (AIF): A computational architecture that decouples interaction-independent components (user-side, item-side) from real-time prediction, allowing them to be precomputed asynchronously. [3]
• Interaction-independent components: Model computations that operate entirely within a single user or item, independent of the specific user-item pair (e.g., user profile processing, item feature embeddings). [3]
• Online Asynchronous Inference: Executing user-side feature fetching and internal network forward computation in parallel with the candidate retrieval stage to reduce pre-ranking latency. [3]
• Implement online asynchronous inference for user-side computations (parallel with retrieval) and nearline asynchronous inference for item-side computations (triggered by updates) to optimize for feature freshness and efficiency. [2]
• Employ Locality Sensitive Hashing (LSH) for approximating similarity computations in long-term user behavior modeling, drastically reducing computational complexity and enabling longer sequences. [2]
• Pre-cache user-side components of cross-features (e.g., SIM-hard subsequences) in parallel with retrieval to mitigate latency bottlenecks from remote feature access and parsing. [2]
• Ensure model version consistency across asynchronous and real-time stages using mechanisms like hashed keys for user-side features and synchronously updated index tables for item-side features. [2]
• Utilize Bridge Embedding Approximation (BEA) to efficiently expand the dimensionality of asynchronously inferred user/item vectors, enhancing model expressiveness without significant computational overhead. [1]

Why we think this paper is great for you:
This research on difference-aware user modeling for LLM personalization is highly relevant to creating more nuanced and effective travel personalization and recommendation systems for you. It offers insights into leveraging user data for tailored travel experiences.

Abstract
Large Language Models (LLMs) are increasingly integrated into users' daily lives, driving a growing demand for personalized outputs. Prior work has primarily leveraged a user's own history, often overlooking inter-user differences that are critical for effective personalization. While recent methods have attempted to model such differences, their feature extraction processes typically rely on fixed dimensions and quick, intuitive inference (System-1 thinking), limiting both the coverage and granularity of captured user differences. To address these limitations, we propose Difference-aware Reasoning Personalization (DRP), a framework that reconstructs the difference extraction mechanism by leveraging inference scaling to enhance LLM personalization. DRP autonomously identifies relevant difference feature dimensions and generates structured definitions and descriptions, enabling slow, deliberate reasoning (System-2 thinking) over user differences. Experiments on personalized review generation demonstrate that DRP consistently outperforms baseline methods across multiple metrics.

Why we think this paper is great for you:
This paper explores how LLMs can enhance geographic item recommendations, which is directly applicable to improving personalized travel recommendations and local travel search experiences for you. It leverages modern AI for more semantic understanding in travel.

Abstract
Recent advances in Large Language Models (LLMs) have enhanced text-based recommendation by enriching traditional ID-based methods with semantic generalization capabilities. Text-based methods typically encode item textual information via prompt design and generate discrete semantic IDs through item tokenization. However, in domain-specific tasks such as local-life services, simply injecting location information into prompts fails to capture fine-grained spatial characteristics and real-world distance awareness among items. To address this, we propose LGSID, an LLM-Aligned Geographic Item Tokenization Framework for Local-life Recommendation. This framework consists of two key components: (1) RL-based Geographic LLM Alignment, and (2) Hierarchical Geographic Item Tokenization. In the RL-based alignment module, we initially train a list-wise reward model to capture real-world spatial relationships among items. We then introduce a novel G-DPO algorithm that uses pre-trained reward model to inject generalized spatial knowledge and collaborative signals into LLMs while preserving their semantic understanding. Furthermore, we propose a hierarchical geographic item tokenization strategy, where primary tokens are derived from discrete spatial and content attributes, and residual tokens are refined using the aligned LLM's geographic representation vectors. Extensive experiments on real-world Kuaishou industry datasets show that LGSID consistently outperforms state-of-the-art discriminative and generative recommendation models. Ablation studies, visualizations, and case studies further validate its effectiveness.

Why we think this paper is great for you:
This research on identifying travel modes from GPS data is highly relevant to understanding travel behavior, which can inform personalized travel experiences and planning tools for you. It provides foundational insights into how people move.

Abstract
Travel mode identification (TMI) from GPS trajectories is critical for urban intelligence, but is hampered by the high cost of annotation, leading to severe label scarcity. Prevailing semi-supervised learning (SSL) methods are ill-suited for this task, as they suffer from catastrophic confirmation bias and ignore the intrinsic data manifold. We propose ST-ProC, a novel graph-prototypical multi-objective SSL framework to address these limitations. Our framework synergizes a graph-prototypical core with foundational SSL Support. The core exploits the data manifold via graph regularization, prototypical anchoring, and a novel, margin-aware pseudo-labeling strategy to actively reject noise. This core is supported and stabilized by foundational contrastive and teacher-student consistency losses, ensuring high-quality representations and robust optimization. ST-ProC outperforms all baselines by a significant margin, demonstrating its efficacy in real-world sparse-label settings, with a performance boost of 21.5% over state-of-the-art methods like FixMatch.

Why we think this paper is great for you:
This work on trajectory retrieval is pertinent to understanding and categorizing travel patterns, which can enhance your ability to personalize travel recommendations and improve travel search functionalities. It offers a lightweight method for handling spatiotemporal data.

Abstract
Trajectory similarity retrieval is an important part of spatiotemporal data mining, however, existing methods have the following limitations: traditional metrics are computationally expensive, while learning-based methods suffer from substantial training costs and potential instability. This paper addresses these problems by proposing \textbf{Geo}metric \textbf{P}rototype \textbf{T}rajectory \textbf{H}ashing (GeoPTH), a novel, lightweight, and non-learning framework for efficient category-based trajectory retrieval. GeoPTH constructs data-dependent hash functions by using representative trajectory prototypes, i.e., small point sets preserving geometric characteristics, as anchors. The hashing process is efficient, which involves mapping a new trajectory to its closest prototype via a robust, \textit{Hausdorff} metric. Extensive experiments show that GeoPTH's retrieval accuracy is highly competitive with both traditional metrics and state-of-the-art learning methods, and it significantly outperforms binary codes generated through simple binarization of the learned embeddings. Critically, GeoPTH consistently outperforms all competitors in terms of efficiency. Our work demonstrates that a lightweight, prototype-centric approach offers a practical and powerful alternative, achieving an exceptional retrieval performance and computational efficiency.

Why we think this paper is great for you:
This paper explores sustainable destination management, which is crucial for the long-term health of the travel industry and effective travel planning. It offers insights into balancing economic and environmental factors in travel.

Abstract
Overtourism poses severe challenges to popular destinations worldwide, threatening natural environments and local communities. This paper develops a decision-making model integrating system dynamics with multi-objective evolutionary algorithms (NSGA-II) to balance economic returns, environmental protection, and social satisfaction. We collect multi-source data from 2008-2024 including visitor arrivals (up to 3.1M), government revenue/expenditure (up to $10.3M), glacier retreat (220-350 ft), CO2 emissions (77K-105K tons), and social satisfaction (0.29-0.48), and establish a dynamic system with four modules: tourist behavior, government finance, environmental evolution, and social well-being. We optimize three objectives via NSGA-II: cumulative net revenue, final environmental index, and final social satisfaction. Experiments on Juneau show optimal solutions yield net revenue up to $1.64B with environmental index 0.93 and social satisfaction 0.86. Extending to Iceland reveals Pareto fronts spanning revenues $150M-$200M, environment indices up to 0.92, and social satisfaction above 0.80. Sobol and Morris sensitivity analyses indicate carbon fees and price elasticity account for over 60% of environmental outcome variance, while capacity limits explain around 90% of net revenue variability. Scenario simulations demonstrate how capacity limits and dynamic pricing on crowded attractions, combined with marketing and infrastructure investment in lesser-known sites, mitigate congestion and enhance sustainability. This work contributes: (i) an integrated system-dynamics and NSGA-II framework for sustainable tourism management; (ii) demonstrated portability via case studies on Juneau and Iceland; and (iii) global sensitivity analysis highlighting influential policy levers for decision makers.

Why we think this paper is great for you:
Understanding how systems recognize places is fundamental to developing more intelligent travel planning and navigation tools for you. This review provides a broad perspective on a core aspect of travel intelligence.

Abstract
Place recognition, the ability to identify previously visited locations, is critical for both biological navigation and autonomous systems. This review synthesizes findings from robotic systems, animal studies, and human research to explore how different systems encode and recall place. We examine the computational and representational strategies employed across artificial systems, animals, and humans, highlighting convergent solutions such as topological mapping, cue integration, and memory management. Animal systems reveal evolved mechanisms for multimodal navigation and environmental adaptation, while human studies provide unique insights into semantic place concepts, cultural influences, and introspective capabilities. Artificial systems showcase scalable architectures and data-driven models. We propose a unifying set of concepts by which to consider and develop place recognition mechanisms and identify key challenges such as generalization, robustness, and environmental variability. This review aims to foster innovations in artificial localization by connecting future developments in artificial place recognition systems to insights from both animal navigation research and human spatial cognition studies.