We consider assortment packs of apparel items which are sold in multiple sizes. Our goal is to jointly optimize the design and allocation of these assortment packs so that the needs of each store in a network for each apparel size can be met by allocating a specific quantity of each pack configuration. We frame the problem as a Bi-Linear Integer Program and present an ensemble optimization method to take advantage of this bi-linearity. This ensemble framework is flexible and can be applied to many variations of this and similar problems. We then evaluate the performance of our method on 10 benchmark datasets from the Target Corporation, which we make public for interested readers to develop their own solutions.
Large-scale e-commerce search must surface a broad set of items from a vast catalog, ranging from bestselling products to new, trending, or seasonal items. Modern systems therefore rely on multiple specialized retrieval channels to surface products, each designed to satisfy a specific objective. A key challenge is how to effectively merge documents from these heterogeneous channels into a single ranked list under strict latency constraints while optimizing for business KPIs such as user conversion. Rank-based fusion methods such as Reciprocal Rank Fusion (RRF) and Weighted Interleaving rely on fixed global channel weights and treat channels independently, failing to account for query-specific channel utility and cross-channel interactions. We observe that multi-channel fusion can be reformulated as a query-dependent learning-to-rank problem over heterogeneous candidate sources. In this paper, we propose a unified ranking model that learns to merge and rank documents from multiple retrieval channels. We formulate the problem as a channel-aware learning-to-rank task that jointly optimizes clicks, add-to-carts, and purchases while incorporating channel-specific objectives. We further incorporate recent user behavioral signals to capture short-term intent shifts that are critical for improving conversion in multi-channel ranking. Our online A/B experiments show that the proposed approach outperforms rank-based fusion methods, leading to a +2.85% improvement in user conversion. The model satisfies production latency requirements, achieving a p95 latency of under 50 ms, and is deployed on Target.com.
Toddlers and older siblings ( M age = 49.34 months) from 145 two-parent, mother-father families (85% European American, 4.9% African American, 3.5 % Asian American, 3.2% Hispanic) participated in a longitudinal investigation when toddlers were 18, 24, and 36 months old. Sibling sharing and the older siblings’ management of the interaction was observed in a laboratory-based fishing game, and mothers and fathers reported on both the toddlers’ and older siblings’ conscience (affective discomfort, moral regulation) at each timepoint. There were significant increases in both siblings’ sharing and older siblings’ management from 18 to 24 months, whereas toddlers’ sharing increased and older siblings’ management decreased from 24 to 36 months. There were stable individual differences in toddlers’ and older siblings’ sharing, and older siblings’ management from 18 to 24 months, but not from 24 to 36 months, suggesting changes in the development of sibling sharing during the second and third years when toddlers become more autonomous contributors to the sharing dynamic and older siblings can step back from managing turn-taking. Reciprocity between older siblings’ and toddlers’ sharing was evident within each time point, but little evidence of bidirectional influence over time. Moral regulation and affective discomfort were highly stable from 18 to 36 months for both siblings, but only the older siblings’ moral regulation at 18 months predicted their sharing at 24 months. Results support a constructivist-interactionist perspective of moral development by showing that sharing develops gradually over time through a set progressive social interactions with a sibling.
To address the difficulty faced by university faculty and students in obtaining useful information from massive campus data, this paper proposes an intelligent campus question-and-answer (Q&A) system based on dynamic retrieval-augmented generation (RAG) technology, using campus administrative knowledge as the data source. The system integrates large language models (LLMs) with domain-specific professional knowledge, leveraging the Campus All-in-One project as a foundation. It constructs a campus knowledge base that includes administrative guides, frequently asked questions, and regulatory documents as an external data corpus. By applying the Infinity database, designed specifically for dynamic RAG applications, and employing prompt engineering, the model’s ability to generate accurate and context-aware answers is enhanced. Through this dynamic RAG-based approach tailored for the education domain, the system provides users with interactive access to a wide range of campus administrative information, helping to resolve common issues, simplify inquiry processes for teachers and students, and reduce the workload of campus management.
Agentic vision-language models increasingly act through extended interactions, but most evaluations still focus on single-image, single-turn correctness. We introduce AMIGO (Agentic Multi-Image Grounding Oracle Benchmark), a long-horizon benchmark for hidden-target identification over galleries of visually similar images. In AMIGO, the oracle privately selects a target image, and the model must recover it by asking a sequence of attribute-focused Yes/No/Unsure questions under a strict protocol that penalizes invalid actions with Skip. This setting stresses (i) question selection under uncertainty, (ii) consistent constraint tracking across turns, and (iii) fine-grained discrimination as evidence accumulates. AMIGO also supports controlled oracle imperfections to probe robustness and verification behavior under inconsistent feedback. We instantiate AMIGO with Guess My Preferred Dress task and report metrics covering both outcomes and interaction quality, including identification success, evidence verification, efficiency, protocol compliance, noise tolerance, and trajectory-level diagnostics.
