candidate filtering

This paper studies the optimal sequential selection of hidden-quality candidates by agents in online labor markets, where agents observe noisy ordinal rankings and binary availability signals (“available/busy”). Although busyness positively correlates with candidate quality, it reduces matching feasibility—introducing a trade-off between value and availability. Using game-theoretic modeling, Bayesian decision analysis, and signaling mechanism design, the paper characterizes agents’ equilibrium sequential selection strategies. It identifies a counterintuitive result: improving ranking accuracy can *reduce* social welfare—by amplifying agents’ tendency to misselect high-quality but busy candidates and increasing the likelihood that lower-ranked yet available candidates are overlooked. This challenges the conventional wisdom that “more accurate rankings are always better,” and provides theoretical foundations for designing platform signaling schemes and optimizing matching mechanisms.

To address insufficient diversity and incomplete coverage of user preferences in multi-generator re-ranking, this paper proposes a comprehensiveness-driven collaborative re-ranking framework. Methodologically, we first formally define and quantify “list comprehensiveness,” then formulate a joint optimization objective balancing preference alignment and comprehensiveness maximization; we further design a learnable complementarity assessment module to enable automatic generator discovery and collaborative scheduling. Our contributions are threefold: (1) the first comprehensiveness metric and optimization paradigm tailored for re-ranking; (2) a learnable modeling mechanism for generator complementarity; and (3) significant improvements in NDCG (+2.1%) and CTR (+1.8%) on two public benchmarks and online A/B tests, empirically validating the framework’s effectiveness in enhancing both recommendation quality and coverage breadth.

This paper studies preference learning from choice feedback over a dynamic item set, aiming to identify either the optimal item or the complete preference ranking with minimal samples and high confidence. We propose two algorithms—Nested Elimination (NE) and Nested Partitioning (NP)—that provide the first non-asymptotic, instance-dependent sample complexity bounds for arbitrary strict preference structures; NE achieves asymptotic optimality in the information-theoretic worst case, while NP attains constant-factor optimality. Our approach integrates multi-dimensional random walk modeling, divide-and-conquer strategies, and an information-theoretic analysis framework. Rigorous theoretical analysis is complemented by extensive experiments on both synthetic and real-world datasets, demonstrating superior efficiency and robustness. The core contribution is the establishment of the first algorithmic framework for dynamic preference learning that simultaneously ensures practical applicability and theoretical optimality.

Existing performance ranking methods in entity evaluation struggle to simultaneously satisfy application-specific preferences and theoretical rigor. Method: This paper establishes the first axiomatic, verifiable general theory framework for performance ranking. Grounded in probability theory and order theory, it formally defines core concepts—including performance objects, satisfaction, and importance—and introduces a performance order satisfying axioms such as ranking consistency, along with constructive procedures for deriving such orders. It further proposes a novel parameterized family of universal ranking scores that unifies classical metrics (e.g., accuracy, recall, F1-score) and rigorously proves that several widely used metrics—including precision—violate the ranking consistency axiom. Contribution/Results: The framework provides the first mathematically rigorous yet practically flexible foundation for performance evaluation in computer vision and machine learning, explicitly characterizing the validity boundaries and intrinsic limitations of reliable ranking metrics.

This paper addresses the challenge of heterogeneous retrieval requirements across multi-task RAG agents by proposing the first unified search engine supporting joint offline learning and online adaptation. Methodologically, it introduces an iterative utility maximization algorithm grounded in the Expectation-Maximization (EM) framework, which jointly models multi-agent feedback to optimize personalized retrieval ranking (Learning to Rank). Key contributions include: (1) the first application of the EM framework to RAG retrieval optimization, enabling joint calibration across heterogeneous LLMs and RAG strategies; and (2) a novel offline–online co-adaptation mechanism that balances system stability with real-time responsiveness. Evaluated on 18 RAG models from the KILT benchmark, our approach achieves statistically significant improvements over strong baselines. Ablation studies validate the efficacy of each component and confirm consistent, agent-specific retrieval gains.

Existing generative recommender systems suffer from a disconnect between semantic ID (SID) construction and personalized ranking objectives, which limits retrieval performance. This work proposes DIG, a novel framework that unifies ranking and retrieval through the lens of tokenization for the first time: it embeds a tokenizer within a discriminative ranking model and trains the entire system end-to-end, leveraging user-item cross features to guide codebook boundary optimization. Additionally, a user-to-token (u2t) distillation module is introduced to enable efficient inference. By design, the ranking model inherently acquires retrieval capabilities, leading to significant improvements across ranking, retrieval, and joint tasks on three public benchmarks and two industrial datasets.

This work addresses the challenges of non-unique solutions and noise-induced temporary infeasibility in structured ranking and selection problems by proposing a unified framework, ENDS. The framework integrates answer-level acceptance sets, a constrained generalized likelihood ratio stopping rule, and a novel answer–trap decomposition mechanism, yielding a max-max-min eigenvalue characterization and a general information-directed sampling principle. By dynamically constructing acceptance sets, explicitly detecting traps, and incorporating cost-aware sampling, ENDS is broadly applicable to diverse settings such as multi-fidelity ranking and Condorcet winner identification. Empirical results demonstrate that the method achieves superior performance across a range of pure exploration tasks, confirming its generality and practical utility.

Existing generative retrieval methods in recommendation systems rely solely on next-token prediction, which struggles to capture the hierarchical structure of user preferences and the deep interactions between items and behavioral sequences. This work proposes RankGR, the first approach to integrate listwise direct preference optimization (Listwise DPO) into generative retrieval. RankGR employs a two-stage collaborative mechanism: an Initial Assessment Phase (IAP) for coarse candidate filtering, followed by a Refinement Scoring Phase (RSP) that leverages a lightweight scoring module for fine-grained ranking. This design balances modeling depth with inference efficiency, enabling high-concurrency real-time deployment. Experiments demonstrate that RankGR significantly improves offline metrics across multiple academic and industrial datasets and delivers substantial online gains in Taobao’s “Guess You Like” scenario, stably supporting nearly 10,000 queries per second.

Traditional scoring functions exhibit inherent theoretical limitations in balancing utility and fairness for ranking tasks. This work rigorously proves, for the first time, that such functions cannot span the entire Pareto frontier of utility–fairness trade-offs. The universality of this limitation is demonstrated through counterexamples constructed across multiple settings, including deterministic versus stochastic scenarios and single-query versus multi-query contexts. To address this gap, the paper introduces a semi-greedy post-processing algorithm that efficiently approximates the ideal solution within a general formal fairness framework. Experimental results show that the proposed method substantially outperforms existing scoring mechanisms and achieves performance close to that of exhaustive post-processing within computationally feasible bounds.

Rankings are highly sensitive to minor data perturbations, which can significantly undermine decision reliability; however, existing approaches lack a quantitative measure of the local impact of perturbations on individual items. To address this gap, this work introduces the notion of “local stability” to characterize how infinitesimal changes to an item’s value affect its rank, and incorporates a dense-region tolerance mechanism that permits rank swaps among nearby items within a local neighborhood. We propose the sampling-based LStability algorithm together with the Detect-Dense-Region method, leveraging concentration inequalities and optimization techniques to enable efficient, scalable stability assessment with theoretical approximation guarantees. Experimental evaluations and case studies demonstrate the effectiveness and practical utility of the proposed framework.