Belhassen Bayar

Yue Zhou, Xiaobo Guo, Belhassen Bayar et al.

Long-term conversational agents face a fundamental scalability challenge as interactions extend over time: repeatedly processing entire conversation histories becomes computationally prohibitive. Current approaches attempt to solve this through memory frameworks that predominantly fragment conversations into isolated embeddings or graph representations and retrieve relevant ones in a RAG style. While computationally efficient, these methods often treat memory formation minimally and fail to capture the subtlety and coherence of human memory. We introduce Amory, a working memory framework that actively constructs structured memory representations through enhancing agentic reasoning during offline time. Amory organizes conversational fragments into episodic narratives, consolidates memories with momentum, and semanticizes peripheral facts into semantic memory. At retrieval time, the system employs coherence-driven reasoning over narrative structures. Evaluated on the LOCOMO benchmark for long-term reasoning, Amory achieves considerable improvements over previous state-of-the-art, with performance comparable to full context reasoning while reducing response time by 50%. Analysis shows that momentum-aware consolidation significantly enhances response quality, while coherence-driven retrieval provides superior memory coverage compared to embedding-based approaches.

Phanideep Gampa, Farnoosh Javadi, Belhassen Bayar et al.

Various data imbalances that naturally arise in a multi-territory personalized recommender system can lead to a significant item bias for globally prevalent items. A locally popular item can be overshadowed by a globally prevalent item. Moreover, users' viewership patterns/statistics can drastically change from one geographic location to another which may suggest to learn specific user embeddings. In this paper, we propose a multi-task learning (MTL) technique, along with an adaptive upsampling method to reduce popularity bias in multi-territory recommendations. Our proposed framework is designed to enrich training examples with active users representation through upsampling, and capable of learning geographic-based user embeddings by leveraging MTL. Through experiments, we demonstrate the effectiveness of our framework in multiple territories compared to a baseline not incorporating our proposed techniques.~Noticeably, we show improved relative gain of up to $65.27\%$ in PR-AUC metric. A case study is presented to demonstrate the advantages of our methods in attenuating the popularity bias of global items.

Belhassen Bayar, Phanideep Gampa, Ainur Yessenalina et al.

Current multi-armed bandit approaches in recommender systems (RS) have focused more on devising effective exploration techniques, while not adequately addressing common exploitation challenges related to distributional changes and item cannibalization. Little work exists to guide the design of robust bandit frameworks that can address these frequent challenges in RS. In this paper, we propose a new design principles to (i) make bandit models robust to time-variant metadata signals, (ii) less prone to item cannibalization, and (iii) prevent their weights fluctuating due to data sparsity. Through a series of experiments, we systematically examine the influence of several important bandit design choices. We demonstrate the advantage of our proposed design principles at making bandit models robust to dynamic behavioral changes through in-depth analyses. Noticeably, we show improved relative gain compared to a baseline bandit model not incorporating our design choices of up to $11.88\%$ and $44.85\%$, respectively in ROC-AUC and PR-AUC. Case studies about fairness in recommending specific popular and unpopular titles are presented, to demonstrate the robustness of our proposed design at addressing popularity biases.

Happy Mittal, Aniket Chakrabarti, Belhassen Bayar et al.

We propose a practical instant question answering (QA) system on product pages of ecommerce services, where for each user query, relevant community question answer (CQA) pairs are retrieved. User queries and CQA pairs differ significantly in language characteristics making relevance learning difficult. Our proposed transformer-based model learns a robust relevance function by jointly learning unified syntactic and semantic representations without the need for human labeled data. This is achieved by distantly supervising our model by distilling from predictions of a syntactic matching system on user queries and simultaneously training with CQA pairs. Training with CQA pairs helps our model learning semantic QA relevance and distant supervision enables learning of syntactic features as well as the nuances of user querying language. Additionally, our model encodes queries and candidate responses independently allowing offline candidate embedding generation thereby minimizing the need for real-time transformer model execution. Consequently, our framework is able to scale to large e-commerce QA traffic. Extensive evaluation on user queries shows that our framework significantly outperforms both syntactic and semantic baselines in offline as well as large scale online A/B setups of a popular e-commerce service.