MambaDATG: Domain-Adaptive Tri-Plane-Gated Pre-training for 3D Abdominal Segmentation

Y. Shen* , D. Jiang* , Zihan Liang* , Y. Wei, W. Yang, S. Wang, et al. | Accepted at ICASSP 2026

• Proposed MambaDATG, a tri-plane-gated selective SSM pre-training that fuses axial, coronal, and sagittal scans via voxel gating—capturing directional anisotropy with linear-time efficiency and zero inference overhead.

• Introduced a domain-adaptive MIM stage adapting high-level layers on unlabeled CTs, improving small-organ segmentation (e.g., +1.8 Dice on BTCV) and revealing the tri-plane gate as the key performance driver.

Causal Representation Learning from Multimodal Clinical Records under Non-Random Modality Missingness

Zihan Liang* , Ziwen Pan*, Ruoxuan Xiong | Accepted at EMNLP 2025 Main

• Proposed CRL-MMNAR, a causal multimodal framework that treats modality missing-not-at-random as signal—combining missingness-aware fusion (gated by observation patterns) with cross-modal reconstruction + contrastive learning, and a multitask predictor with a cross-fitted rectifier to correct observation-pattern bias.

• Demonstrated consistent gains on MIMIC-IV and eICU, including AUC 0.8687→0.9824 for ICU admission (+13.1%) and 0.7989→0.8657 for readmission (+8.4%) on MIMIC-IV, and readmission AUC to 0.9294 (+13.8%) on eICU, with lower Brier scores indicating better calibration.

CareLab at #SMM4H-HeaRD 2025: Insomnia Detection and Food Safety Event Extraction with Domain-Aware Transformers

Zihan Liang*, Ziwen Pan*, Sumon Kanti Dey, Azra Ismail | Accepted at AAAI ICWSM 2025

• Ranked 1st on SMM4H–HeaRD 2025 Task 5, developing a RoBERTa + GPT-4–augmented system with classweighted training and ensembling, achieving F1 = 0.958 and revealing the limits of rule-based span extraction in clinical and food-safety texts.

Dynamic Policy Design for Autonomous Taxi Adoption: A Hierarchical Game-Theoretic Framework

Zihan Liang, Ziwen Pan | Accepted at TRB Annual Meeting 2026

• Formulated a three-tier dynamic Stackelberg game modeling interactions among government, platform, and workers in autonomous taxi adoption, deriving equilibrium policies that balance innovation incentives with social welfare.

• Under review at Transportation Research Record.

Manuscripts

Learning Dynamic Representations and Policies from Multimodal Clinical Time-Series with Informative Missingness

Zihan Liang* , Ziwen Pan*, Ruoxuan Xiong | ACL 2026 Under Review | Oct. 2025 – Present

• Developed an MNAR-aware multimodal state encoder for offline clinical RL by extending GRU-D with explicit missingness features (time gaps, cumulative counts, missing rates, windowed frequencies) and fusing step-sparse clinical text via cross-attention + adaptive gating, with reconstruction to preserve missingness signals.

• Designed an action-conditioned latent belief dynamics (VAE) module (with a multi-step credit assignment guarantee) and integrated it with Implicit Q-Learning for sepsis treatment optimization, achieving AUROC 0.876 and +20.3% policy improvement over clinician behavior (largest gains in high-severity strata).

Behavioral Deviation as Deliberative Signal: Hierarchical State Decomposition for Sequential Multimodal Recommendation

Ziwen Pan*, Zihan Liang* , Ruoxuan Xiong | ACL 2026 Under Review | Oct. 2025 – Present

• Developed HSD-SMR, a hierarchical debiasing layer that corrects sequential recommender states via group-stratified temporal statistics plus a deviation-vector–driven, reliability-gated individual update.

• Engineered a 16-D behavioral deviation feature set and multi-task training (BPR ranking + rating/modality/content auxiliaries) on a causal Transformer encoder, achieving 3–7% gains over SOTA across three Amazon categories.

DART: Mitigating Harm Drift in Difference-Aware LLMs via Distill-Audit-Repair Training

Ziwen Pan*, Zihan Liang* , Jad Kabbara, Ali Emami | ACL 2026 Under Review | Oct. 2025 – Present

• Proposed DART for difference-awareness classification, combining label-conditioned teacher rationale distillation, LoRA-based fine-tuning, and a structured inference-time explanation policy to control rationale content.

• Built a harm-drift detection + targeted repair loop using paired baseline vs. distilled generations, toxicity-delta screening (\tau=0.01) and LLM-as-judge severity stratification, then severity-weighted repair, improving Llama-3-8B accuracy 39.0% → 68.8% while reducing drift cases 435 → 119 (−72.6%).

Adaptive Spatiotemporal Graph Neural Networks with Trend-Aware Prediction and Validation-Gated Calibration for Traffic Flow Forecasting

Zihan Liang* , Ziwen Pan*, Shuyang Yu | IEEE T-ITS Under Review | May 2025 – Present

• Proposed an adaptive spatiotemporal GNN that fuses distance- and correlation-based graphs with dual temporal–spatial encoders and a trend-aware head, achieving state-of-the-art accuracy on PeMS benchmarks.

• Designed a validation-gated calibration pipeline with hour-conditional quantile mapping and AR(1) residual correction, applied only when improving MAPE/MAE to ensure reliable and interpretable post-processing.

Gated Dynamic Local-Global Attention for Sentiment Analysis: Enhancing Context Awareness in Short Texts

Ziwen Pan*, Zihan Liang* | Applied Intelligence Under Review | Jun 2024 – Present

• Developed a Gated Dynamic Local–Global Attention (GDLGA) architecture that combines BERT-based global context modeling with sentiment-adaptive dynamic local attention, enabling window sizes to expand or contract based on token-level emotional intensity.

• Introduced a token-level gating mechanism and multi-granularity pooling to fuse hierarchical global features with fine-grained local representations, achieving more stable, calibrated, and context-aware sentiment classification in short texts.

Collaborative Research Projects

Research Assistant at Collective Action & Research for Equity (CARE) Lab, Emory University (Sept. 2024 - Present)

• Analyzed nearly 1,000 user data points and developed an interactive dashboard for MakerGhat, an India-based organization, utilizing JavaScript, Python, and HTML for seamless data visualization.

• Maintaining a WhatsApp Education Bot using Python and Twilio for MakerGhat, enabling over 15,000 teachers across India to document lesson plans and teaching activities, impacting 600,000+ students and supporting 10,000+ educational projects.

• Developed an offline-first Android app for classroom session recording and metadata management, integrating Google Sheets validation, offline caching, and secure Drive synchronization.

Research Assistant at Polymath Jr REU Summer Program (Jun. 2025 – Aug. 2025)

• Developed and analyzed mathematical models using ordinary differential equations (ODEs) to study the community transmission of Clostridioides difficile (C. difficile).

• Collaborated with a team to identify optimal intervention strategies for controlling the spread of C. difficile in community settings.

• Assisted in modifying and implementing code in MATLAB and R to simulate disease dynamics and evaluate mitigation strategies.