Zixin Zhao

Patrick Yung Kang Lee, Jessica Y. Bo, Zixin Zhao et al.

Individuals are turning to increasingly anthropomorphic, general-purpose chatbots for AI companionship, rather than roleplay-specific platforms. However, not much is known about how individuals perceive and conduct their relationships with general-purpose chatbots. We analyzed semi-structured interviews (n=13), survey responses (n=43), and community discussions on Reddit (41k+ posts and comments) to triangulate the internal dynamics, external influences, and steering strategies that shape AI companion relationships. We learned that individuals conceptualize their companions based on an interplay of their beliefs about the companion's own agency and the autonomy permitted by the platform, how they pursue interactions with the companion, and the perceived initiatives that the companion takes. In combination with the external factors that affect relationship dynamics, particularly model updates that can derail companion behaviour and stability, individuals make use of different types of steering strategies to preserve their relationship, for example, by setting behavioural instructions or porting to other AI platforms. We discuss implications for accountability and transparency in AI systems, where emotional connection competes with broader product objectives and safety constraints.

Chen Gao, Zixin Zhao, Lv Shao et al.

In e-commerce, Trigger-Induced Recommendation (TIR), recommending items after a user clicks a trigger, is an important task. However, modern platforms rely on a continuous stream of diverse and short-lived promotional scenarios (e.g., for Black Friday), creating a significant challenge. Existing methods are less effective here: they either fall into a trigger-dependency trap, recommending overly similar items, or a data-hungry trap, requiring long-term stable data for intent modeling that these ephemeral scenarios cannot provide. To address these limitations, we propose the Collaborative Contrastive Network (CCN), a general and robust framework that approaches the problem from a different perspective. Instead of modeling ambiguous entry intent, CCN learns a user's context-specific preferences by treating the user-trigger pair as a unique condition. It achieves this via a novel contrastive learning scheme, using the collaborative feedback of co-click/co-non-click as a positive signal and mono-click as a negative signal to structure the item representation latent space. To prove its real-world generality, CCN is trained on a heterogeneous dataset spanning over a dozen different scenarios from an entire year, and the online A/B test is conducted in a completely new, unseen scenario on Taobao, where CCN boosts CTR by 12.3\% and order volume by 12.7\%, demonstrating its effectiveness and generalization.

Chen Gao, Zixin Zhao, Lv Shao et al.

Click-Through Rate (CTR) prediction has long been dominated by discriminative paradigms that optimize local decision boundaries within candidate-specific subspaces. However, these models often fail to capture the global joint distribution and the continuous structural evolution of user intent across all-domain movelines. While generative approaches attempt to model global transition patterns, existing methods suffer from discretization-induced information collapse by remapping nuanced e-commerce signals into discrete linguistic or categorical spaces, failing to preserve the topological fidelity of interest trajectories. To overcome these limitations, we propose a novel generative pre-training paradigm that models user intent as a continuous evolutionary trajectory on a high-dimensional latent interest manifold, termed the Next Interest Flow (NIF). We introduce kinematic constraints to govern this flow: Interest Diversity is achieved via tangent space decomposition, while Evolution Velocity ensures trajectory smoothness through geodesic regularization. To bridge the objective mismatch between generative pre-training and discriminative fine-tuning, we propose a bidirectional alignment strategy to synchronize semantic spaces. Furthermore, we develop a Temporal Sequential Pairwise (TSP) mechanism to instill temporal causality within the discriminative framework. We present the All-domain Moveline Evolution Network (AMEN), a unified framework implementing this pipeline. Extensive experiments on a 6.7-billion instance industrial dataset and online A/B tests on Taobao validate AMEN's superiority, achieving +0.87pt AUC gain and +11.6\% CTCVR lift.

Chen Gao, Zixin Zhao, Lv Shao et al.

E-commerce app users exhibit behaviors that are inherently logically consistent. A series of multi-scenario user behaviors interconnect to form the scene-level all-domain user moveline, which ultimately reveals the user's true intention. Traditional CTR prediction methods typically focus on the item-level interaction between the target item and the historically interacted items. However, the scene-level interaction between the target item and the user moveline remains underexplored. There are two challenges when modeling the interaction with preceding all-domain user moveline: (i) Heterogeneity between items and scenes: Unlike traditional user behavior sequences that utilize items as carriers, the user moveline utilizes scenes as carriers. The heterogeneity between items and scenes complicates the process of aligning interactions within a unified representation space. (ii) Temporal misalignment of linked scene-level and item-level behaviors: In the preceding user moveline with a fixed sampling length, certain critical scene-level behaviors are closely linked to subsequent item-level behaviors. However, it is impossible to establish a complete temporal alignment that clearly identifies which specific scene-level behaviors correspond to which item-level behaviors. To address these challenges and pioneer modeling user intent from the perspective of the all-domain moveline, we propose All-domain Moveline Evolution Network (AMEN). AMEN not only transfers interactions between items and scenes to homogeneous representation spaces, but also introduces a Temporal Sequential Pairwise (TSP) mechanism to understand the nuanced associations between scene-level and item-level behaviors, ensuring that the all-domain user moveline differentially influences CTR predictions for user's favored and unfavored items. Online A/B testing demonstrates that our method achieves a +11.6% increase in CTCVR.