Two Publications at AAAI 2026
[AAAI 2026] GeoGen — A Two-stage Coarse-to-Fine Framework for Fine-grained Synthetic Location-based Social Network Trajectory Generation
This paper is led by his PhD student Rongchao Xu. This work introduces GeoGen, a two-stage coarse-to-fine framework for large-scale Location-based Social Network (LBSN) check-in trajectory generation. In the first stage, we reconstruct spatially continuous, temporally regular latent movement sequences from the original LBSN check-in trajectories and then design a Sparsity-aware Spatio-temporal Diffusion model (S2TDiff) with an efficient denoising network to learn their underlying behavioral patterns. In the second stage, we design Coarse2FineNet, a Transformer-based Seq2Seq architecture equipped with a dynamic context fusion mechanism in the encoder and a multi-task hybrid-head decoder, which generates fine-grained LBSN trajectories based on coarse-grained latent movement sequences by modeling semantic relevance and behavioral uncertainty. Dr. Wang is the corresponding author of this paper.
[AAAI 2026] TrustEnergy — A Unified Framework for Accurate and Reliable User-level Energy Usage Prediction
This paper is led by his PhD student Dahai Yu. This work introduces a unified framework named TrustEnergy for accurate and reliable energy usage prediction. TrustEnergy consists of two key novel components: (i) a Hierarchical Spatiotemporal Representation module (HSTR), which is capable of efficiently capturing both micro user-level and macro region-level patterns based on a new memory augmented spatiotemporal graph neural network (MASTGNN), and (ii) a Sequential Conformalized Quantile Regression module (SCQR), which is a distribution-agnostic uncertainty quantification method that dynamically adjusts uncertainty bounds to provide valid prediction intervals over time for reliable prediction without making strong assumptions about the underlying data distributions. Dr. Wang is the corresponding author of this paper.
KDD 2026: Uncertainty-Aware Energy Consumption Prediction
[KDD 2026] EnergyMamba — An Uncertainty-Aware Graph-Enhanced Selective State Space Model for Energy Consumption Prediction
This paper is led by his PhD student Dahai Yu. This work introduces EnergyMamba, an uncertainty-aware spatiotemporal learning framework comprising: (i) GE-Mamba, a novel architecture that injects spatial context learned from grid topology into a bidirectional Mamba, organized within a U-Net structure; and (ii) Adaptive Sequential Conformalized Quantile Regression (AS-CQR), a distribution-free uncertainty quantification method with locally adaptive normalization and online feedback calibration for reliable prediction under non-stationary conditions. Dr. Wang is the corresponding author of this paper.
IMWUT/UbiComp 2026: Synthesizing Human Activity Traces
[IMWUT 2026] SynHAT — A Two-stage Coarse-to-Fine Diffusion Framework for Synthesizing Human Activity Traces
This paper is led by his PhD student Rongchao Xu. This work introduces SynHAT, an efficient two-stage coarse-to-fine diffusion framework for Human Activity Traces (HAT) synthesis. A shared Human Activity Diffusion (HADiff) architecture is proposed and applied in both stages with different inputs. It includes a novel Latent Spatio-Temporal UNet (LST-UNet) for denoising through dual Drift–Jitter branches for jointly modeling smooth spatial transitions and temporal variations. In stage 1, we convert the real HATs into coarse-grained latent ST traces for HADiff training to output the synthetic latent traces as the input of the second stage. In stage 2, we design a three-step pipeline composed of BPEM, Fine-HADiff, and Semantic Alignment to capture more precise spatio-temporal dynamics, enhance temporal fidelity, and produce the final fine-grained synthetic HATs. Dr. Wang is the corresponding author of this paper.
IJCAI 2026: Healthcare Facility Visit Prediction
[IJCAI 2026] HealthMamba — An Uncertainty-aware Spatiotemporal Graph State Space Model for Effective and Reliable Healthcare Facility Visit Prediction
This paper is led by his PhD student Dahai Yu. This work introduces HealthMamba, an uncertainty-aware spatiotemporal framework for accurate and reliable healthcare facility visit prediction. HealthMamba comprises three key components: (i) a Unified Spatiotemporal Context Encoder that fuses heterogeneous static and dynamic information, (ii) a novel Graph State Space Model called GraphMamba for hierarchical spatiotemporal modeling, and (iii) a comprehensive uncertainty quantification module integrating three uncertainty quantification mechanisms for reliable prediction. We evaluate HealthMamba on four large-scale real-world datasets from California, New York, Texas, and Florida. Results show HealthMamba achieves around 6.0% improvement in prediction accuracy and 3.5% improvement in uncertainty quantification over state-of-the-art baselines. Dr. Wang is the corresponding author of this paper.
ICRA 2026: Human-Robot Collaboration for Urban Services
[ICRA 2026] UrbanHuRo — A Two-Layer Human-Robot Collaboration Framework for the Joint Optimization of Heterogeneous Urban Services
This paper is led by his PhD students Tonmoy Dey and Lin Jiang. This work introduces UrbanHuRo, a two-layer human-robot collaboration framework for joint optimization of heterogeneous urban services, demonstrated through the examples of crowdsourced delivery and urban sensing. There are two innovative designs in UrbanHuRo, i.e., (i) a scalable distributed MapReduce-based K-Submodular maximization module for efficient order dispatch and (ii) a deep submodular reward reinforcement learning algorithm for sensing route planning. Dr. Wang is the corresponding author of this paper.
ICLR 2026: Human-centered Assessment
[ICLR 2026] AtC — Aggregate-then-Calibrate for Human-centered Assessment
This paper is led by a Ph.D. student under his supervision. This work introduces Aggregate-then-Calibrate (AtC), a two-stage framework that combines complementary sources. Stage-1 aggregates heterogeneous comparative judgments into a consensus ranking using a rank-aggregation model that accounts for annotator reliability. Stage-2 calibrates any predictive model’s scores by an isotonic projection onto the order, enforcing ordinal consistency while preserving as much of the model’s quantitative information as possible. Dr. Wang is the corresponding author of this paper.
VLDB 2026: Link Discovery in Billion-Scale Heterogeneous Graphs
[VLDB 2026] BiLink — Bidirectional Meta-paths for Link Discovery in Billion-Scale Heterogeneous Graphs
This paper is led by a Ph.D. student under his supervision. This work introduces bidirectional meta-path, a unified path schema for bidirectional paths connecting node pairs. Based on this, we present BiLink, a framework for link discovery in billion-scale heterogeneous graphs. Specifically, BiLink represents all meta-paths in a unified table format and implements bidirectional meta-path sampling in distributed systems with uniform table operations. It also includes BiPathsNN, a model that represents path instances as embedding sequences and jointly encodes them via the Transformer encoder for link prediction. Dr. Wang is the corresponding author of this paper.