Generative recommendation (GR) has shown great promise in industrial applications, particularly for candidate generation and end-to-end recommendations. However, existing GR training paradigms suffer from two fundamental mismatches with real-world deployment requirements. First, they optimize for point-wise prediction of a single ground-truth item, whereas practical systems must produce a diverse, high-value set of candidates. Second, they treat all user interactions as equally informative, ignoring their inherent differences in utility. Although reward-based fine-tuning offers a partial remedy, it often lacks token-level supervision. To address these challenges, we reformulate GR as a sequential set-generation problem and propose GFlowGR, a GFlowNet-based fine-tuning framework that explicitly aligns generation probabilities with item-level utilities. GFlowGR comprises three tightly integrated components, each addressing a key limitation of conventional fine-tuning: a trajectory sampler that constructs training trajectories from candidate sets to enable set-wise learning, a behavior-aware reward model that quantifies item utility to support value-aware optimization, and a GFlowNet objective that provides token-level supervision. Extensive experiments on three real-world datasets with two representative LLM-based GR backbones show consistent and significant improvements over strong baselines, validating the effectiveness of our approach. For real-world deployment, GFlowGR has been integrated into Taobao 's search advertising businesses, delivering a 0.4% relative improvement in annual revenue since its launch in mid-2025, corresponding to billion-level monetary gains. Code is available at https://github.com/Applied-Machine-Learning-Lab/SIGIR26_GFlowGR.

Generative Recommendation (GR), powered by Large Language Models (LLMs), represents a promising new paradigm for industrial recommender systems. However, their practical application is severely hindered by high inference latency, making them infeasible for high-throughput, real-time services and limiting their overall business impact. While Speculative Decoding (SD) has been proposed to accelerate the autoregressive generation process, existing implementations introduce new bottlenecks: they typically require separate draft models and model-based verifiers, which require additional training and increase latency overhead. In this paper, we address these challenges with NEZHA, a novel architecture that achieves hyperspeed decoding for GR systems without sacrificing recommendation quality. Specifically, NEZHA integrates a nimble autoregressive draft head directly into the primary model, enabling efficient self-drafting. This design, combined with a specialized input prompt structure, preserves the integrity of sequence-to-sequence generation. Furthermore, to tackle the critical problem of hallucination—a major source of performance degradation—we introduce an efficient, model-free verifier based on a hash set. We demonstrate the effectiveness of NEZHA through extensive experiments on public datasets and have successfully deployed the system on Taobao since October 2025, achieving 1.2% business improvement, translating to billion-level advertising revenue and serving hundreds of millions of daily active users. The code is available at https://github.com/Applied-Machine-Learning- Lab/WWW2026_NEZHA.

Abstract In CRYPTO 2019, Gohr made a pioneering attempt and successfully applied deep learning to the differential cryptanalysis against NSA block cipher Speck 32/64, achieving higher accuracy than the pure differential distinguishers. By its very nature, mining effective features in data plays a crucial role in data-driven deep learning. In this paper, in addition to considering the integrity of the information from the training data of the ciphertext pair, domain knowledge about the structure of differential cryptanalysis is also considered into the training process of deep learning to improve the performance. Meanwhile, taking the performance of the differential-neural distinguisher of Simon 32/64 as an entry point, we investigate the impact of input difference on the performance of the hybrid distinguishers to choose the proper input difference. Eventually, we improve the accuracy of the neural distinguishers of Simon 32/64, Simon 64/128, Simeck 32/64 and Simeck 64/128. We also obtain related-key differential-based neural distinguishers on round-reduced versions of Simon 32/64, Simon 64/128, Simeck 32/64 and Simeck 64/128 for the first time.

Abstract In this work, we investigate how the choice of the key-expansion algorithm and its interaction with the round function affect the resistance of Simon-like ciphers against rotational-XOR cryptanalysis. We observe that, among the key-expansion algorithms we consider, Simon is most resistant, while Simeck is much less so. Implications on lightweight ciphers design are discussed and open questions are proposed.

Abstract Rotational‐XOR (RX) cryptanalysis is a cryptanalytic method aimed at finding distinguishable statistical properties in Addition‐Rotation‐XOR‐C ciphers, that is, ciphers that can be described only by using modular addition, cyclic rotation, XOR and the injection of constants. In this study, we extend RX‐cryptanalysis to AND‐RX ciphers, a similar design paradigm where the modular addition is replaced by vectorial bitwise AND; such ciphers include the block cipher families Simon and Simeck. We analyse the propagation of RX‐differences through AND‐RX rounds and develop a closed form formula for their expected probability. Inspired by the MILP verification model proposed by Sadeghi et al., we develop a SAT/SMT model for searching compatible RX‐characteristics in Simon‐like ciphers, that is, that there is at least one right pair of messages/keys to satisfy the RK‐characteristics. To the best of our knowledge, this is the first model that takes the RX‐difference transitions and value transitions simultaneously into account in Simon‐like ciphers. Meanwhile, we investigate how the choice of the round constants affects the resistance of Simon‐like ciphers against RX‐cryptanalysis. Finally, we show how to use an RX‐distinguisher for a key recovery attack. Evaluating our model we find compatible RX‐characteristics of up to 20, 27 and 34 rounds with respective probabilities of 2 −26 , 2 −44 and 2 −56 for versions of Simeck with block sizes of 32, 48 and 64 bits, respectively, for large classes of weak keys in the related‐key model. In most cases, these are the longest published distinguishers for the respective variants of Simeck. In the case of Simon, we present compatible RX‐characteristics for round‐reduced versions of all 10 instances. We observe that for equal block and key sizes, the RX‐distinguishers cover fewer rounds in Simon than in Simeck. Concluding the paper, we present a key recovery attack on Simeck 64 reduced to 28 rounds using a 23‐round RX‐characteristic.

In knowledge-sharing communities like Stack Overflow (SO), users can post questions, give answers and choose one answer as an accepted answer. The accepted answers will be important references for users when they encounter similar questions. Essentially, posting questions and giving answers is an interactive process occurring among community users, and choosing accepted answers is actually a decision-making process involving multiple factors. Previous works examined the impact on this decision process from the user, question and answer viewpoints. Social interactions between the questioners and answerers, although being popular according to our pre-analysis, have never been considered as a factor that can influence the decisions. To fill this gap, this paper first proposes a comprehensive answer acceptance model that integrates the answer features established by social interactions as well as information of users, questions and answers. We then divide social interactions into two stages and propose a method to calculate the relationship between the questioner and the answerer by analyzing these social interactions. Finally, we investigate the influence of social interactions for the acceptance of answers by performing logistic regression analysis. The results reveal several findings: (1) social-based features explain 16.6 % of the variance explained together, indicating that social interactions have significant and important effects on the acceptance of answers; (2) social interactions that occur after the answer is posted are more influential than these occur before the answer is posted. Based on the findings, we further conduct an online study of 132 SO users, and the respondents report that social interactions have a greater impact on the acceptance of answers than other judgments of answers such as upvotes, downvotes and not accepting answers.

Stack Overflow is the most popular programming question and answer community that continuously receives a large number of questions every day. To ensure the quality of questions, the community grants privileges for the moderators and a group of experienced users to review the quality of questions and close the low-quality ones (e.g. duplicate or irrelevant questions). The review process is a typical crowdsourcing job that relies on users’ volunteer participation, and the current practices of closing questions in Stack Overflow face two aspects of challenges: (1) an obvious increase in both the absolute number and the percentage of “closed” questions; (2) a considerable decrease in participation willingness of experienced users to close questions. In order to solve the problem, we present a novel model of user willingness for reviewing and voting questions by incorporating four types of user activity history, including questions, answers, comments and votes of closing questions. Then we propose an automatic recommendation method based on the model to assign experienced users proper questions, to utilize the forces of them to close questions. The evaluation shows that the successful recommendation probability in the top 5, top 10, top 20, top 30, top 40, top 50 users are 48.23%, 58.93%, 68.83%, 74.27%, 78.13% and 81%, respectively.