Recently, multi-agent debate-based inference methods have been widely adopted for inappropriate utterance detection to aggregate diverse perspectives and improve explainability. However, performing real-time interactions for every input incurs substantial latency and cost, and may amplify error entrenchment as early misjudgments are repeatedly justified. In this paper, we propose the Reflective Memory and Interventional Context Alignment (REMICA) framework. Specifically, REM stores multi-agent predictions and rationales as offline memory for efficient retrieval-based inference. During this process, ICA enforces alignment constraints to ensure consistency between rationales and predictions. Experiments on four datasets with three commercial LLMs show that REMICA reduces the average inference time relative to online multi-agent inference while improving the average F1 score, with GPT-5.1-mini achieving a relative average F1 improvement of +2.594% and an average inference-time reduction of 7.80× in the best-performing setting. These results suggest that the proposed framework can help mitigate the trade-off between accuracy and latency by reusing offline memory and reducing the risk of error entrenchment. Furthermore, the memory and code developed in this paper are publicly available at https://github.com/wndudwkd003/REMICA.

Argumentative explainable artificial intelligence employs argumentation theory to explain the mechanisms of machine learning. Previous approaches for explaining deep learning models collectively compressed layers via clustering. However, this resulted in accumulated information loss across layers, thereby degrading the fidelity of explanations. We propose online activation value-aware clustering and aggregation, a compression algorithm that preserves the inference structure of the original neural network with greater fidelity. The proposed method sequentially compresses each layer, immediately recalculates activation values following compression, and rectifies inter-layer information loss using a singular-value-scaled ridge alignment approach. To evaluate the effectiveness of the proposed method, we introduce four novel quantitative metrics. Input-output fidelity and structural fidelity measure how accurately the compressed model preserves the original model predictions and internal activations. Input-output perturbation consistency and structural perturbation consistency assess the similarity of the changes induced by Gaussian-perturbed input data. Experiments on three benchmark datasets (Breast Cancer, California Housing, and HIGGS) demonstrate that our method achieves performance improvements ranging from 12.9% to 53.7% across the four metrics, demonstrating significantly higher explanation fidelity than existing approaches.

Sequential recommendation aims to predict a user's next interaction based on their historical behavior.While recent models have achieved remarkable success, they often overlook time intervals between interactions or rely on fixed thresholds for session segmentation, which can lead to suboptimal results.To address these limitations, several approaches incorporate time intervals via relative positional embeddings or session segmentation based on fixed thresholds.However, these methods are highly sensitive to threshold selection and are prone to inaccurate segmentation.Inspired by these challenges, we propose TiSRec, a Time Interval-wise Segmentation framework that dynamically divides user sequences into Local Preference Blocks (LPBs) by selecting significant time intervals.TiSRec captures evolving user preferences through intra-block and inter-block encoders.Experiments on four real-world datasets demonstrate that TiSRec consistently outperforms state-of-the-art methods, and ablation studies confirm the effectiveness of LPBbased modeling.

In Binary Product Quantization (BPQ), where product quantization is applied to binary data, the traditional k-majority method is used for clustering, with centroids determined based on Hamming distance and majority vote for each bit. However, this approach often leads to a degradation in clustering quality, negatively impacting BPQ's performance. To address these challenges, we introduce Binary-to-Real-and-Back K-Means (BRB-KMeans), a novel method that initially transforms binary data into real-valued vectors, performs k-means clustering on these vectors, and then converts the generated centroids back into binary data. This innovative approach significantly enhances clustering quality by leveraging the high clustering quality of k-means in the real-valued vector space, thereby facilitating future quantization for binary data. Through extensive experiments, we demonstrate that BRB-KMeans significantly enhances clustering quality and overall BPQ performance, notably outperforming traditional methods.