This paper studies the problem of Continuous-Time Counterfactual Quantile Learning (CT-CQL) for risk-sensitive policy optimization. In many real-world applications such as patient blood pressure monitoring, financial market analysis, and autonomous driving, data is high-frequency and continuously evolving. However, most existing causal inference methods focus on expectation-based or discrete-time counterfactual reasoning, which fail to capture fine-grained temporal dynamics. As a result, policies optimized under these frameworks may overlook critical risks—e.g., a treatment policy with good average outcomes may still expose patients to life-threatening episodes. To overcome these limitations, we propose CT-CQL, a framework built upon a novel identification theory and featuring three key components: (1) modeling full counterfactual outcome distributions via Stochastic Differential Equations (SDEs) governed by the Fokker–Planck Equation (FPE); (2) enhancing robustness through a minimax objective that minimizes FPE residuals under adversarial perturbations; and (3) mitigating confounding bias using a double-robust AIPW loss. CT-CQL enables robust policy optimization by identifying optimal intervention strategies that maximize expected utility while adhering to real-world budget and safety constraints. Experiments on widely-used benchmarks and the real-world MIMIC-III dataset demonstrate the validation and superiority of the proposed method. The project is available at: https://github.com/Eliza-YiHe/CT-CQL/

Generalizing causal effects from Randomized Controlled Trials (RCTs) to target populations across diverse environments is of significant practical importance, as RCTs are often costly and logistically complex to conduct. A key challenge is environmental shift, defined as changes in the distribution and availability of covariates between source and target environments. A common approach addressing this challenge is to identify a separating set–covariates that govern both treatment effect heterogeneity and environmental differences–and combine RCT samples with target populations matched on this set. However, this approach assumes that the separating set is fully observed and shared across datasets, an assumption often violated in practice. We propose a novel Two-Stage Doubly Robust (2SDR) method that relaxes this assumption by allowing the separating set to be observed in only one of the two datasets. 2SDR leverages shadow variables to impute missing components of the separating set and generalize treatment effects across environments in a two-stage procedure. We show the identification of causal effects in target environments under 2SDR and demonstrate its effectiveness through extensive experiments on both synthetic and real-world datasets.

Given a text query, the task of scene text retrieval aims at searching and localizing all the text instances that are contained in an image gallery. The state-of-the-art method learns a cross-modal similarity between the query text and the detected text regions in natural images to facilitate retrieval. However, this cross-modal approach still cannot well bridge the heterogeneity gap between the text and image modalities. In this paper, we propose a new paradigm that converts the task into a single-modality retrieval problem. Unlike previous works that rely on character recognition or embedding, we directly leverage pictorial information by rendering query text into images to learn the glyph feature of each character, which can be utilized to capture the similarity between query and scene text images. With the extracted visual features, we devise a synthetic label image guided feature alignment mechanism that is robust to different scene text styles and layouts. The modules of glyph feature learning, text instance detection, and visual matching are jointly trained in an end-to-end framework. Experimental results show that our proposed paradigm achieves the best performance in multiple benchmark datasets. As a side product, our method can also be easily generalized to support text queries with unseen characters or languages in a zero-shot manner.