arXiv subjects

Hui Gao

Publications and source records attributed to Hui Gao.

At least 19 recordsLinked to original sources

arXiv2026

STAIR: Semantic-Temporal Automaton for Interpretable Reasoning in Temporal Question Answering

By leveraging large-scale pretraining, LLMs can interpret diverse temporal expressions and question formulations without task-specific training. However, existing prompt-based neuro-symbolic systems continue to rely on LLMs for both semantic interpretation and exact temporal inference. Consequently, discrete decisions regarding intervals, time anchors, and ordered states remain vulnerable to probabilistic errors and difficult to verify. We present STAIR, a \textbf{S}emantic-\textbf{T}emporal \textbf{A}utomaton for \textbf{I}nterpretable \textbf{R}easoning. STAIR separates semantic interpretation from precise temporal inference: an answer-free LLM adapter maps complex question formulations to normalized temporal intents, while a deterministic temporal automaton with finite control and guarded transitions executes the corresponding policies over canonicalized evidence. Following a rule-first design, STAIR resolves standard questions without invoking an LLM and applies semantic adaptation only when the rule path fails to produce an executable intent. This approach reduces free-form reasoning, making temporal decisions verifiable and interpretable. Specifically, guarded execution supports precise point-time containment and before/after selection, while semantic adaptation handles non-exact intervals and time-anchored queries. Across the TimeQA-Easy, TimeQA-Hard, TempReason-L2, and TempReason-L3 datasets, STAIR consistently outperforms strong baselines in the TQA task using matched model settings, achieving average F1 improvements of 16.57\% and 3.10\% when utilizing the Qwen2.5-7B and GPT-4o-mini models, respectively. Furthermore, ablations and diagnostic analyses demonstrate that STAIR excels at handling both boundary-sensitive and order-sensitive queries, while its guarded execution and semantic adaptation ensure precise point-time reasoning and inexact intervals, respectively.

cs.CL↗

arXiv2026

DSETA: A Dual-Stage Continual Learning Framework for Travel Time Prediction in Dynamic Traffic Environments

Estimated Time of Arrival (ETA) prediction is a core component of intelligent transportation systems. As traffic congestion patterns become increasingly dynamic in large cities, maintaining high prediction accuracy poses a major challenge for ride-hailing platforms. Existing methods either fail to adapt to irregular traffic patterns and sudden congestion, or suffer from new distributions without disentangling long-term trends from short-term fluctuations, thereby degrading model performance in real-world scenarios. To address this challenge, we propose DSETA, an incrementally updated Dual-Stage ETA prediction framework. Specifically, the continual learning process is divided into \textit{inter-day} and \textit{intra-day} stages. We first design the \textit{intra-day} learning stage, which relies entirely on real-time data to enable dynamic adaptation to short-term traffic patterns caused by events like holidays or accidents. Next, we develop the \textit{inter-day} learning stage, which leverages aggregated historical data from a short time window to capture knowledge of long-term distribution shifts, such as seasonal trends and traffic network evolution. Subsequently, to prevent catastrophic forgetting and preserve knowledge of regular patterns, we explore a \textit{Historical Traffic Knowledge Consolidation} module. Finally, we validate DSETA's effectiveness and robustness through extensive offline and online experiments conducted on real-world datasets from DiDi's platform. Online A/B tests across three major cities including Beijing, Wuhan, and Xi'an consistently demonstrated performance gains, achieving MAE reductions of 6.62\%, 0.73\%, and 2.40\% respectively. This framework has been successfully deployed in DiDi's production environment, processing hundreds of millions of daily requests and validating its strong performance in industrial applications.

cs.LG↗

arXiv2026

de Rham theory and locally analytic vectors

Let $K_\infty/K$ be a $p$-adic Lie extension of a $p$-adic field $K$. We study the subring of pro-analytic vectors in the de Rham period ring $\mathbf{B}_{\mathrm{dR}}^+(K_\infty)$. We show that the pro-analytic subring admits a Galois-equivariant isomorphism with a formal power series ring $\widehat{K}_{\infty}^{\mathrm{la}} [[t_{K_\infty}]]$ if and only if $K_\infty$ satisfies a certain orientability condition, which says that the $\widehat{K}_\infty$-level Sen operator admits a Galois-equivariant $\mathbf{B}_{\mathrm{dR}}^+$-lift. A key input is the vanishing of higher locally analytic vectors of $\widehat{K}_\infty$-representations. As an application, we show that the lifted Sen operator induces regular connections on pro-analytic vectors of $\mathbf{B}_{\mathrm{dR}}^+$-representations, and can be used to compute Galois cohomology.

math.NT↗

arXiv2026

Proximity Features: Privacy-Compliant Cold-Start Personalization at Airbnb

Personalization in two-sided marketplaces relies heavily on user-level features, yet for platforms with infrequent, high-consideration purchases, a large fraction of users lack sufficient history for effective recommendation, spanning both paid and organic channels. At Airbnb, a substantial share of search requests comes from logged-out or first-time users, with this challenge especially pronounced on paid-channel landing pages, leaving traditional user-level features unavailable for a large fraction of traffic. Privacy regulations and increasing restrictions on third-party cookies further limit identifier-based tracking for non-essential use cases. This paper introduces Proximity Features, a privacy-compliant feature system that groups users by geographic proximity using geo-IP data and an adaptive clustering algorithm, producing aggregated user-level signals for groups of approximately 1,000 nearby users without requiring a persistent individual identifier at inference time. Privacy is preserved by design: the pipeline operates on consented, aggregated data only within consent-gated privacy controls. The system is deployed in production at Airbnb, serving multiple surfaces including marketing landing pages and destination recommendation, with engagement emails integration under way. Online A/B experiments demonstrate statistically significant lifts in bookings, with the largest gains observed among users with absent or stale history.

cs.LG↗

arXiv2026

SAMA: Semantic Anchor-aligned Augmentation for Unified Low-Resource Multimodal Information Extraction

Multimodal Information Extraction (MIE)-covering tasks such as Multimodal Named Entity Recognition (MNER), Relation Extraction (MRE), and Event Extraction (MEE)-is essential for understanding multimedia content but remains constrained by severe data scarcity. Although data augmentation is a promising remedy, existing approaches are impeded by coarse cross-modal alignment and fragmented, task-specific designs that fail to exploit shared semantic knowledge. To overcome these limitations, we introduce Semantic Anchor-aligned Multimodal Augmentation (SAMA), a unified framework for generating high-fidelity, task-aware synthetic data. SAMA constructs structured semantic anchors from ground-truth labels to guide a Collaborative Multi-Experts Multimodal Large Language Model (CME-MLLM), which integrates a Universal Adapter for shared semantics with Task-Specific Adapters to produce diverse yet constraint-compliant textual samples. For image synthesis, SAMA employs an Anchor-Preserving Diffusion mechanism that uses anchor-weighted prompts and latent conditioning to maintain critical semantic anchors while diversifying visual contexts. To eliminate the need for manual verification, SAMA further introduces a Dual-Constraint Filtering module that selects synthetic samples based on both cross-modal consistency and anchor fidelity. Extensive experiments across benchmark datasets for MNER, MRE, and MEE demonstrate that SAMA consistently outperforms state-of-the-art augmentation baselines under both fully supervised and low-resource settings, underscoring its versatility, robustness, and effectiveness.

cs.CV↗

arXiv2026

JourneyFormer: Encoding Airbnb Guest Journey with Sequence Modeling

Sequence modeling has become increasingly popular in recommendation and ranking algorithms, owing to its capacity to model users' historical behaviors and infer user intentions. Despite its theoretical simplicity, the practical deployment of a sequence model in production is non-trivial due to complexity of the sequence and sparse labels. For example, in Airbnb, guest sequences are often long, exploratory and complex, and we focus on booking labels, which are sparse. As such, we are often required to make various design decisions regarding data and modeling to strike a balance between effectiveness and scalability. This work delved into these production challenges and deployed JourneyFormer, a sequence modeling solution for search ranking at Airbnb. We detail crucial design considerations, covering aspects such as guest event selection, ID embeddings, model architecture, and label attribution. Additionally, we describe several tailored strategies to accelerate model training and inference. JourneyFormer has been successfully deployed within Airbnb's production, where its effectiveness and impact have been evidenced not only by improved offline ranking metrics but also by significant gains in key business metrics through online A/B testing across 2 production surfaces.

cs.LG↗

arXiv2026

Artificial Intelligence for Subsurface Imaging Understanding: A Decade Review of Challenges, Methods, Benchmarks, and Outlook

Subsurface imaging interpretation bridges observed geophysical data and quantitative geological models, supporting hydrocarbon exploration, CO2 storage assessment, and geohazard monitoring. Over the past decade, machine learning and deep learning have substantially reshaped interpretation workflows. This review synthesizes the 2015-2025 literature across four tasks: structural interpretation, geobody identification, seismic facies analysis, and property estimation, tracing the field's evolution from classical machine learning through deep learning to emerging domain foundation models, and how these tasks couple within a single interpretation system. The task remains fundamentally different from other AI applications, facing ambiguous signals, interpretive non-uniqueness, sparse semantics, unfixed target locations, and scarce reliable annotations. We synthesize three defining challenges: interpretation under complex geological conditions, cross-survey semantic generalization under low information density, and the absence of reliable benchmarks. Addressing them will hinge on integrating human expertise, physical constraints, and geological priors into training and inference, and on treating uncertainty quantification as an intrinsic model output. We outline a forward-looking agenda: unified, jointly modelled interpretation systems with cross-task consistency; priors evolving from physics toward language and multimodal supervision; end-to-end uncertainty propagation; human-AI collaboration and agent-orchestrated workflows; and a more rigorous evaluation science supported by an AI-ready data ecosystem. The review is accompanied by an open benchmark resource (CIG-Bench), covering fault segmentation, relative geologic time estimation, geobody segmentation, and property modeling, with synthetic datasets, pretrained baselines, and quantitative evaluation: https://douyimin.github.io/CIG-bench

physics.geo-ph↗

arXiv2026

Implicit Structural Modeling via Generative Diffusion Frameworks

Implicit structural modeling can support understanding subsurface spatial configurations, revealing patterns of geological evolution, and enabling quantitative simulation of geological processes, thereby offering substantial scientific and engineering value. Conventional approaches formulate it as an optimization problem or framework interpolation to fit a continuous scalar field, whereas machine learning methods typically adopt discriminative regression to directly predict implicit models. However, in complex scenarios involving fault intersections, branching, and thrust nappes, these methods still struggle to maintain topological consistency and kinematic plausibility. In this work, we develop an implicit structural modeling approach based on diffusion models. We construct a set of training data through a simulation based synthesis pipeline and design a dedicated encoder for conditional injection, allowing the conditional branch to converge rapidly while effectively reinforcing the input conditional priors throughout the diffusion process, thereby more stably propagating structural constraints. We then inject these conditional features into a backbone network pretrained on large scale natural images to enable conditional training of the diffusion model. Although our synthetic data include only a relatively stylized normal fault system, experiments demonstrate strong generalization, enabling the model to effectively handle diverse complex structural types such as strike slip faults and intricate flower fault systems. More importantly, even in challenging thrust nappe settings where the scalar field becomes non monotonic and exhibits abrupt depth discontinuities, the model can still generate reliable implicit structural models.

physics.geo-ph↗

arXiv2026

Pretrain-to-alignment learning paradigm to improve geophysical AI applicability under scarce field labels and synthetic-to-field gaps: A case study of relative geologic time estimation in global shelf-edge clinothems

Artificial intelligence (AI) has been increasingly applied to various geophysical scenarios, yet its practical deployment remains limited by scarce field labels, pronounced synthetic-to-field domain gaps, and insufficient physical consistency under complex and variable field conditions. To address these challenges, we propose a pretrain-to-alignment learning paradigm that systematically integrates self-supervised pretraining, synthetic supervision, prior-driven refinement, and domain-adaptation fine-tuning into a unified progressive learning workflow. In this paradigm, geophysical AI models are developed through sequential stages that progressively build field-relevant representations, task-specific mapping capability, field consistency, and target-specific adaptability. We validate this paradigm using cross-survey relative geologic time (RGT) estimation in global shelf-edge clinothems as a representative case study. Results from 3,000 field datasets spanning multiple sedimentary basins demonstrate that the proposed paradigm achieves accurate, robust, and well-generalized performance across diverse field surveys, while significantly improving fine-scale stratigraphic and structural details. More broadly, this study provides a practical methodological reference for a broader range of geophysical AI tasks, such as interpretation, regression, and inversion problems.

physics.geo-ph↗

arXiv2026

Learning Stratigraphically Consistent Relative Geologic Time from 3D Seismic Data via Sinusoidal Mapping

Relative Geologic Time (RGT) estimation from seismic data is a cornerstone of subsurface structural modeling, depositional evolution analysis, and reservoir characterization, supporting horizon correlation and depositional system reconstruction. Yet accurate RGT estimation remains challenging: RGT is intrinsically a topologically constrained continuous field, in which local errors readily propagate globally and distort the overall result. Conventional methods rely heavily on priors, attribute extraction, and manual interaction, leading to cumbersome workflows. Existing deep-learning approaches mostly use a regression formulation with pixel-wise MSE/MAE losses, which struggle to capture thin horizons and fail to model the stratigraphic semantics of the RGT field, yielding limited generalization and unstable ordering across diverse structural and depositional settings. We propose RGT-Est, a deep-learning framework that transfers the optimization target from the topologically constrained continuous field into a differentiable sinusoidal space, which explicitly encodes the periodic stratigraphic semantics of RGT and alleviates over-smoothing of fine horizons. Pointwise, perceptual, and adversarial losses are jointly imposed in this space to enforce local fidelity, inter-layer consistency, and global structural plausibility, providing both fine-horizon discrimination and global stratigraphic awareness. An optional horizon-guidance module further accepts sparse 2D or 3D horizons as priors. Trained on synthetic data and evaluated on field surveys with densely faulted zones, large unconformities, steeply dipping strata, folded deformations, and clinoforms, RGT-Est achieves state-of-the-art performance among AI-based methods without horizon constraints, and attains substantially higher horizon-correlation accuracy and global topological consistency once sparse priors are incorporated.

physics.geo-ph↗

arXiv2026

Galois representations over convergent de Rham period ring

Let $\mathbf{B}_{\mathrm{dR}}^{+, \dagger} \subset \mathbf{B}_{\mathrm{dR}}^{+}$ be the ``convergent" de Rham period ring which is the (un-completed) stalk at the de Rham point of the Fargues--Fontaine curve. We develop a Tate--Sen formalism to relate Galois representations over $\mathbf{B}_{\mathrm{dR}}^{+, \dagger}$ to regular connections over convergent functions. As a consequence, when the Sen weights (of the mod $t$ reduction) satisfy a $p$-adic non-Liouville condition, Galois cohomology of a $\mathbf{B}_{\mathrm{dR}}^{+, \dagger}$-representation compares to that of its $\mathbf{B}_{\mathrm{dR}}^{+}$-base change, and hence is finite. In addition, restricted to objects whose Sen weights are algebraic numbers, the categories of $\mathbf{B}_{\mathrm{dR}}^{+, \dagger}$-representations and $\mathbf{B}_{\mathrm{dR}}^{+}$-representations are equivalent.

math.NT↗

arXiv2026

Massive-scale unlabeled field and labeled synthetic seismic datasets of global shelf-edge clinothems

Seismic stratigraphic interpretation of shelf-edge clinothems is essential for revealing tectonic evolution, paleoclimate change, depositional dynamic conditions, and hydrocarbon generation and accumulation during basin filling. However, traditional interpretation methods remain labor-intensive, time-consuming, and highly subjective. Although AI-based method offer a potential solution for automated this task, its development has been limited by the scarcity of comprehensive and representative benchmark datasets for shelf-edge clinothems. This limitation primarily arises from limited field data availability, the scarcity of reliable geological labels, and the structural complexity and strong variability of clinothem-dominated systems. To address this gap, we develop a hybrid benchmark dataset through two complementary strategies of field data curation and geological and geophysical forward modeling, ultimately generating 3,000 unlabeled field and 4,000 labeled synthetic seismic data, respectively. We further evaluate several representative baseline deep learning models on these datasets, and the accurate results demonstrate that the curated dataset provides an effective and representative basis for model training, quantitative assessment, and practical application. Finally, we have publicly released this hybrid benchmark dataset (https://doi.org/10.5281/zenodo.18910271) to facilitate the development, validation, and assessment of deep learning methods for automated seismic stratigraphic interpretation.

physics.geo-ph↗

arXiv2026

Applying Embedding-Based Retrieval to Airbnb Search

The goal of Airbnb search is to match guests with the ideal accommodation that fits their travel needs. This is a challenging problem, as popular search locations can have around a hundred thousand available homes, and guests themselves have a wide variety of preferences. Furthermore, the launch of new product features, such as \textit{flexible date search,} significantly increased the number of eligible homes per search query. As such, there is a need for a sophisticated retrieval system which can provide high-quality candidates with low latency in a way that integrates with the overall ranking stack. This paper details our journey to build an efficient and high-quality retrieval system for Airbnb search. We describe the key unique challenges we encountered when implementing an Embedding-Based Retrieval (EBR) system for a two sided marketplace like Airbnb -- such as the dynamic nature of the inventory, a lengthy user funnel with multiple stages, and a variety of product surfaces. We cover unique insights when modeling the retrieval problem, how to build robust evaluation systems, and design choices for online serving. The EBR system was launched to production and powers several use-cases such as regular search, flexible date and promotional emails for marketing campaigns. The system demonstrated statistically-significant improvements in key metrics, such as booking conversion, via A/B testing.

cs.IR↗

arXiv2025

Large Emotional World Model

The world is governed by both physical laws and affective dynamics. Physical laws govern state transitions, while affective dynamics shape human actions, decisions, and interactions. A world model that learns only physical laws can approximate the physical world, but not the human world. In this paper, we introduce human emotion as a key state variable in world models, enabling them to capture both future state transitions and their emotional causes. We first construct Emotion-Why-How (EWH), the first world model dataset centered on emotional state transitions, containing 10,850 emotion-aware transition tuples. Each tuple encodes the pre-state, pre-emotion, action, post-emotion, and post-state, supporting reasoning about why actions occur and how emotions reshape future states. Based on EWH, we propose the Large Emotional World Model (LEWM), which factorizes future prediction into two coupled steps: first predicting the future emotional state from the current context, and then conditioning future world-state prediction on the predicted emotion. Experiments show that LEWM brings consistent gains across world-state prediction, emotion understanding, and general reasoning tasks. It achieves up to 45.72% accuracy improvement on EWH, 3.94% on WorldNet, 17.47% F1 improvement on MELD, and a 6.10% gain on specific MMLU categories. These results demonstrate that incorporating emotion into world models enables more realistic simulation of human-centered environments and expands the predictive understanding of intelligent agents.

cs.CL↗

arXiv2025

Packing spanning arborescences with extra large one

The celebrated Nash-Williams and Tutte's theorem states that a graph $G=(V, E)$ contains $k$ edge disjoint spanning trees if and only if $\nu_{f}(G) \geq k$, where $$\nu_{f}(G):=\min_{|\mathcal{\mathcal{P}}|>1, \text{$\mathcal{P}$ is a partition of $V(G)$}}\frac{|E( \mathcal{P})|}{|\mathcal{P}|-1}.$$ Inspired by the NDT theorem as structural explanations for the fractional part of Nash-Williams' forest decomposition theorem, Fang and Yang extended Nash-Williams and Tutte's theorem and proved that if $\nu_{f}(G) > k+ \frac{d-1}{d}$, then $G$ contains $k$ edge disjoint spanning trees and another forest $F$ with $ |E(F)|> \frac{d-1}{d} (|V(G)|-1)|$, and if $F$ is not a spanning tree, then $F$ has a component with at least $d$ edges. In this paper, we give a digraphic version of their result; however, the mixed graphic version remains open.

math.CO↗

arXiv2025

Fine-Grained DINO Tuning with Dual Supervision for Face Forgery Detection

The proliferation of sophisticated deepfakes poses significant threats to information integrity. While DINOv2 shows promise for detection, existing fine-tuning approaches treat it as generic binary classification, overlooking distinct artifacts inherent to different deepfake methods. To address this, we propose a DeepFake Fine-Grained Adapter (DFF-Adapter) for DINOv2. Our method incorporates lightweight multi-head LoRA modules into every transformer block, enabling efficient backbone adaptation. DFF-Adapter simultaneously addresses authenticity detection and fine-grained manipulation type classification, where classifying forgery methods enhances artifact sensitivity. We introduce a shared branch propagating fine-grained manipulation cues to the authenticity head. This enables multi-task cooperative optimization, explicitly enhancing authenticity discrimination with manipulation-specific knowledge. Utilizing only 3.5M trainable parameters, our parameter-efficient approach achieves detection accuracy comparable to or even surpassing that of current complex state-of-the-art methods.

cs.CV↗

arXiv2025

Brevity is the Soul of Wit: Condensing Code Changes to Improve Commit Message Generation

Commit messages are valuable resources for describing why code changes are committed to repositories in version control systems (e.g., Git). They effectively help developers understand code changes and better perform software maintenance tasks. Unfortunately, developers often neglect to write high-quality commit messages in practice. Therefore, a growing body of work is proposed to generate commit messages automatically. These works all demonstrated that how to organize and represent code changes is vital in generating good commit messages, including the use of fine-grained graphs or embeddings to better represent code changes. In this study, we choose an alternative way to condense code changes before generation, i.e., proposing brief yet concise text templates consisting of the following three parts: (1) summarized code changes, (2) elicited comments, and (3) emphasized code identifiers. Specifically, we first condense code changes by using our proposed templates with the help of a heuristic-based tool named ChangeScribe, and then fine-tune CodeLlama-7B on the pairs of our proposed templates and corresponding commit messages. Our proposed templates better utilize pre-trained language models, while being naturally brief and readable to complement generated commit messages for developers. Our evaluation based on a widely used dataset showed that our approach can outperform six baselines in terms of BLEU-Norm, METEOR, and ROUGE-L, with average improvements of 51.7%, 78.7%, and 62.5%, respectively. The ablation study and human evaluation also provide further insights into the effectiveness of our approach.

cs.SE↗

arXiv2025

SAC-MIL: Spatial-Aware Correlated Multiple Instance Learning for Histopathology Whole Slide Image Classification

We propose Spatial-Aware Correlated Multiple Instance Learning (SAC-MIL) for performing WSI classification. SAC-MIL consists of a positional encoding module to encode position information and a SAC block to perform full instance correlations. The positional encoding module utilizes the instance coordinates within the slide to encode the spatial relationships instead of the instance index in the input WSI sequence. The positional encoding module can also handle the length extrapolation issue where the training and testing sequences have different lengths. The SAC block is an MLP-based method that performs full instance correlation in linear time complexity with respect to the sequence length. Due to the simple structure of MLP, it is easy to deploy since it does not require custom CUDA kernels, compared to Transformer-based methods for WSI classification. SAC-MIL has achieved state-of-the-art performance on the CAMELYON-16, TCGA-LUNG, and TCGA-BRAC datasets. The code will be released upon acceptance.

cs.CV↗