ranking and allocation algorithm design

This paper addresses the fundamental trade-off between relevance and advertising revenue in job ranking on recruitment platforms. We propose a joint preference-aware ranking and position-aware auction mechanism. Methodologically, we integrate causal inference estimation, multi-objective optimization, and mechanism design theory to construct a real-time deployable ranking–auction co-optimization framework that preserves short-term platform revenue while enhancing long-term matching quality. Our key contribution lies in dynamically coupling job seeker preference modeling with ad-slot value estimation, enabling Pareto-improving relevance gains under strict revenue constraints. Empirical results demonstrate that, with less than 1% advertising revenue loss, click-through rate and application conversion rate increase by over 12%, while user retention and overall platform health significantly improve.

In multi-slot sponsored search advertising, budget-based allocation mechanisms induce unfairness among advertisers in terms of impression and conversion distribution. Method: This paper proposes an online traffic allocation framework that jointly optimizes platform efficiency and advertiser fairness. It introduces the Gini index—novelly applied to quantify advertising fairness—and formulates a bi-objective constrained optimization model balancing efficiency and fairness. Departing from conventional budget-driven auction designs, the framework solves an online combinatorial optimization problem with explicit fairness constraints. A lightweight, real-time deployable algorithm is developed to ensure no degradation in overall platform revenue while improving fairness. Results: Extensive experiments on multiple real-world datasets demonstrate that the method consistently outperforms mainstream auction baselines in both efficiency and fairness metrics. It offers theoretical rigor, engineering practicality, and offline evaluability.

In e-commerce advertising, Sponsored Listing Ranking (SLR) must jointly optimize short-term revenue and long-term user experience under stringent real-time latency constraints (<0.1 sec). To address this, we propose the first large-scale online SLR method integrating linear programming relaxation with dual optimization, formulating a scalable constrained mixed-integer program that flexibly incorporates operational constraints—such as inventory availability and fairness—beyond conventional heuristic scoring approaches. Our method enables controllable, verifiable multi-objective optimization, overcoming the rigidity of fixed-weight heuristics in balancing competing objectives. Evaluated over a 19-day field experiment on a leading e-commerce platform (329 million impressions), it significantly improves advertising revenue versus industrial baselines while preserving search relevance. Deployed system-wide in January 2023.

This work addresses exposure inequality, reduced diversity, and regulatory risks in online platform recommendations caused by algorithmic bias toward popular items. We propose FAIR, the first combinatorial item-selection framework explicitly enforcing *pairwise fairness*: it ensures approximately equal exposure probabilities for any pair of items via linear programming. Methodologically, we introduce a provably 1/2-approximation algorithm and a fully polynomial-time approximation scheme (FPTAS), integrating the ellipsoid method, parameterized knapsack approximation, and a dual separation oracle. Experiments on MovieLens and synthetic datasets validate FAIR’s effectiveness, quantify the “fairness cost” (i.e., the trade-off between fairness and utility), and demonstrate its ability to jointly optimize fairness, diversity, and recommendation revenue.

This paper studies the student-school assignment problem under capacity constraints and group-level fairness requirements, jointly optimizing individual utilities (e.g., preference rankings), school enrollment caps, and inter-group fairness—such as across ethnicity or geography—formulated either via concave objective functions or explicit group-wise constraints, and supporting arbitrary covering constraints to capture multi-criteria and ordinal optimization needs. We propose, for the first time, a unified algorithmic framework that integrates convex programming modeling with systematic rounding techniques, yielding tunable randomized or deterministic algorithms. These run in polynomial time and provide controlled trade-offs among utility loss, capacity violations, and fairness deviations. Theoretically, our approach achieves provable approximation guarantees and naturally generalizes to covering constraints and ranking-aware settings. It exhibits strong scalability and practical deployability.

This study addresses the problem of dynamic fair allocation of indivisible goods (or tasks) in an online setting, where items arrive sequentially and must be allocated immediately upon arrival. Under a broad range of models—including normalized and non-normalized utilities as well as identical and general additive utility functions—the work designs constructive online algorithms within the competitive analysis framework, targeting multiple fairness criteria such as EF1 and PROP1. For most settings considered, the paper not only presents algorithms achieving optimal competitive ratios but also establishes matching theoretical upper bounds, thereby substantially expanding the theoretical foundations of online fair division.

This study addresses the problem of jointly selecting items and users in e-commerce marketing campaigns to construct non-overlapping, high-quality promotion groups that maximize matching effectiveness. The task is formalized for the first time as an automatic targeting problem, and a novel combinatorial optimization framework is proposed, integrating constrained spectral co-clustering, greedy local search with pairwise exchanges, and multi-armed bandits to jointly optimize effectiveness, fairness, and scalability. Experimental results on synthetic data, Amazon Reviews, and large-scale real-world commercial datasets demonstrate that the proposed co-clustering approach achieves superior performance in campaign quality, uplift, and fairness, while the multi-armed bandit variant exhibits stronger scalability in ultra-large-scale scenarios.

This work addresses the challenge of reconciling Pareto efficiency with max-min fairness in machine learning training resource allocation under heterogeneous user workload valuations. The authors propose and deploy a market-based GPU quota exchange platform featuring a dynamic pricing mechanism that enables users to explicitly express the value of their tasks, thereby guiding efficient resource allocation. To the best of our knowledge, this system is the first to simultaneously guarantee both Pareto efficiency and max-min fairness in settings with heterogeneous task values, successfully integrating market mechanisms into a large-scale production environment. Deployment within Google demonstrates significant improvements in resource utilization, effective support for diverse business-critical workloads, and the opening of new avenues for scheduling optimization.