Jul 2026· Annual International ACM SIGIR Conference on Research and Development in Information Retrieval· pp. 3647-3651· 0 citations· 17 references
Computer Science
TL;DR
This work is the first to reformulate the recommendation denoising problem into a Partial Label Learning (PLL) task, and innovatively leverages PLL paradigms to directly resolve ambiguous implicit feedback, effectively recovering clean signals from noisy candidate sets.
Abstract
Implicit feedback, such as clicks and browsing behaviors, is ubiquitous in recommender systems as a proxy for user preferences. However, these signals are inherently noisy; interactions such as misclicks, unintended views, or unsatisfactory purchases often introduce false positive patterns that mislead model learning. Existing denoising strategies mainly rely on heuristic ''small-loss'' principles to suppress the influence of high-loss samples. However, this approach creates a fundamental trade-off: by indiscriminately penalizing large-value losses, these methods inadvertently weaken the model's ability to learn ''hard'' true positive interactions, thereby compromising robustness and personalization. By conceptualizing this ambiguity as a ''candidate label set'' that encompasses both true and noisy feedback, we are the first to reformulate the recommendation denoising problem into a Partial Label Learning (PLL) task. This novel perspective allows us to address the fundamental challenge of unreliable pseudo-labels by transforming traditional heuristic-based filtering into a principled label disambiguation process. Specifically, we propose PLLD, a Partial Label Learning-inspired Denoising method. Unlike existing methods that rely on indirect signal filtering, PLLD innovatively leverages PLL paradigms to directly resolve ambiguous implicit feedback, effectively recovering clean signals from noisy candidate sets. Experiments on multiple real-world benchmark datasets demonstrate that PLLD consistently improves ranking performance and robustness under substantial noise.
In recommendation systems, users interact with only a small fraction of a vast item catalog, producing feedback that is both sparse and noisy. This challenges post-training generative recommenders: reward models trained from logged interactions often fail to generalize, while directly optimizing imperfect rewards can lead to reward over-optimization. We propose Exponential reward-weighted fine-tuning (Exp-RSFT), where each logged interaction is weighted by $\exp(r/\lambda)$, avoids this failure by optimizing directly on the logged rewards, with the temperature $\lambda$ regularizing against their noise. We theoretically show that Exp-RSFT's suboptimality decomposes into two costs: a coverage cost arising from limitations of the logging policy and a noise cost from imperfect feedback. The temperature $\lambda$ balances these competing effects, yielding an optimal tradeoff between exploiting high-reward behavior and robustness to noise. Across three public benchmarks and a large-scale industrial dataset, we verify this theoretical prediction: performance follows an inverted-U trend as a function of $\lambda$, while PPO and DPO often over-optimize unreliable reward models and degrade recommendation quality. Exp-RSFT consistently improves ranking performance without requiring online exploration or preference data.
Recommender systems often struggle to balance global preference patterns, local similarity, and complex user-item interactions within a single model. Existing approaches combine traditional collaborative filtering (CF) methods such as Singular Value Decomposition (SVD) and K-Nearest Neighbors (KNN) with gradient boosting models such as XGBoost as meta-learners, yet they often fail to capture complex non-linear interactions. In addition, XGBoost may exhibit limitations in stability, as it can become sensitive to training data and show signs of overfitting in certain scenarios. To address this limitation, this study proposes a stacking ensemble-based CF approach that extends a baseline model by incorporating Neural Collaborative Filtering (NCF) as an additional base learner and employing Random Forest (RF) as the meta-learner. These base models capture complementary characteristics: SVD models global latent factors, KNN captures local similarity, and NCF learns non-linear user-item interactions. Experiments are conducted on the MovieLens 100K dataset. The results demonstrate that augmenting the baseline model with NCF and RF reduces prediction ambiguity and improves system convergence stability. The proposed ensemble achieves the lowest RMSE of 0.9113. Although the baseline model with XGBoost slightly outperforms in MAE, the proposed approach achieves higher Recall and F1-score. These results indicate better overall recommendation quality.
Muhammad Faried Gunawan, Rita Rismala· International Conference on...· 0 citations
Search and recommendation (S&R) are fundamental components of modern commercial platforms, enabling users to access and explore information efficiently. User behaviors in these scenarios reflect different aspects of user intent, providing an opportunity for joint modeling of S&R. However, effectively leveraging search logs to enhance recommendation remains a challenging task. Existing methods often encode S&R histories either jointly or separately; however, they tend to regard all search signals as equally informative, thereby neglecting that many search behaviors can be irrelevant or even detrimental to recommendation performance. In practice, however, search histories frequently contain noisy or outdated behaviors that may introduce spurious correlations and degrade recommendation performance. Motivated by the human decision-making process, where one first identifies recommendation intent and then selectively reasons about relevant search signals, we propose LCR-SER, a latent cross reasoning method for search-enhanced recommendation. LCR-SER first encodes the user's S&R history into a unified latent representation that captures users' global interests. It then performs iterative reasoning in the latent space to dynamically identify informative search signals that are most relevant to the recommendation. To further guide this reasoning process, we introduce contrastive learning to align the reasoning states with the target items. In addition, we employ reinforcement learning to directly optimize ranking-oriented metrics, enabling LCR-SER to refine its reasoning strategy toward improved recommendation performance. Experiments on public datasets demonstrate that LCR-SER consistently outperforms strong baselines, validating the effectiveness of latent reasoning in enhancing search-aware recommendation.
Teng Shi, Weicong Qin, Weijie Yu et al.· Annual International ACM SIG...· 1 citation
User interest modeling is foundational to recommender systems. However, sparse and noisy behaviors make traditional item-level sequence models brittle, especially for new and low-activity users. Furthermore, relying solely on a user's own history limits exploration and reinforces the ''filter bubbles''. To address this, we propose GEM (Generative Enhanced Modeling). GEM shifts the paradigm from self-behavior induction to collective experience migration. Specifically, it constructs LLM-based semantic IDs and embeddings. Grounded in information theory, GEM performs multi-stage denoising at both the user and item levels. This design effectively suppresses reward-driven noise while preserving target-aware signals. We deployed GEM on the Alipay Tab3 video feed. Offline evaluations show significant GAUC gains. Online A/B tests demonstrate a 0.9% lift in watch time alongside stable video views and improved exposure diversity. These results confirm that GEM enhances recommendation quality and successfully broadens user interests.
Li Li, Wei Xu, Yu Cheng et al.· Annual International ACM SIG...· 0 citations
Sequential recommendation models, while powerful, have limited flexibility in responding to immediate user requests, making it difficult to adapt their recommendations to the user's timely interests. Unfortunately, existing user request adaptation methods often incur high computational overhead due to either 1) retraining the entire backbone network or 2) leveraging the inference ability of large language models (a.k.a. prompt engineering), limiting their applicability in large-scale recommendation services. This paper presents Control via Request-Aware Masking for Editing Recommenders (CRAMER), a framework that takes users'natural-language requests to immediately change sequential recommendation models'behavior. Specifically, inspired by the model control theory, CRAMER treats user requests as control signals to modulate frozen backbone parameters through masking, achieving instant adaptation to diverse requests while avoiding costly retraining. Experiments on multiple large-scale benchmark datasets show that CRAMER outperforms four state-of-the-art request-aware baselines across multiple recommendation metrics while achieving minimal overhead. Moreover, the proposed framework exhibits enhanced controllability and cross-domain adaptability, establishing a new paradigm for request-aware sequential recommendation.
Zhiyuan Su, Naihe Feng, Zhen Qin et al.· 0 citations
It is suggested that LLM-derived priors can serve as a practical warm-start mechanism for text-rich bandit recommendation, while also revealing deployment trade-offs.
E. Lee, Oseong Choi, Byungsoo Kang et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
