Multibehavior recommender systems are proficient at constructing precise representations of users and items by leveraging a variety of interaction behaviors, e.g., click, add-to-cart, and purchase. However, they are still facing the following challenges: 1) there are a large number of spurious correlations existing in auxiliary behaviors (i.e., noise unrelated to the target behavior preference); and 2) spurious correlations may be unintentionally introduced and amplified during the representation learning process in traditional multibehavior recommenders. Hence, we propose a novel framework-disentangling spurious correlations (DISC) to measure and disentangle the spurious correlations of the multibehavior recommendation. In particular, to precisely identify the intent representations for each user, we devise a cross-behavior self-attention layer incorporating lightweight graph convolution networks to encode graph nodes under specific behaviors, thereby enabling the model to supervise the subsequent spurious correlation disentanglement. Subsequently, to disentangle the spurious correlations among multiple behaviors, we design a time-sensitive Jaccard coefficient to dynamically measure the users’ spurious correlations. Building on this, we propose a dual mutual information (MI) bound structure to disentangle spurious correlations existing in auxiliary behaviors and transfer genuine correlated semantic information to the target behavior, thus alleviating the data sparsity. Extensive experiments on three real-world datasets demonstrate the consistent improvements obtained by DISC over 11 state-of-the-art baselines by effectively disentangling the spurious correlations.

Recent advancements in single-cell sequencing technologies enable the measurement of multiple modalities in individual cells, offering insights into the transcriptome and regulome in various biological systems and human diseases in an unprecedented resolution. However, effectively using these ultra-high-dimensional and large-scale multiomic data to understand gene regulation remains challenging. Inspired by the success of adapting large language models into the genomics field, we develop scMBERT, a BERT framework-based pre-trained deep learning model using single-cell multiomic data. We showed that scMBERT increases model flexibility and performance in downstream tasks like cell type annotation and batch-effect correction, demonstrating the potential of leveraging multiomic data to improve single-cell genomic data analyses.

The success of large language models has driven interest in developing similar speech processing capabilities. However, a key challenge is the scarcity of high-quality spontaneous speech data, as most existing datasets contain scripted dialogues. To address this, we present a novel pipeline for eliciting and recording natural dialogues and release our dataset with 100+ hours of spontaneous speech. Our approach fosters fluid, natural conversations while encouraging a diverse range of topics and interactive exchanges. Unlike traditional methods, it facilitates genuine interactions, providing a reproducible framework for future data collection. This paper introduces our dataset and methodology, laying the groundwork for addressing the shortage of spontaneous speech data. We plan to expand this dataset in future stages, offering a growing resource for the research community.

In e-commerce, recommender systems help users find suitable products by leveraging diverse behaviors, e.g., view, cart and buy. In recent years, multi-behavior recommender systems have made strides by integrating auxiliary behaviors with purchase histories to deliver high-quality recommendations. However, most existing methods often fail to identify spurious correlation intents within auxiliary behaviors that conflict with users' target intents. Indiscriminately incorporating such correlations into the prediction of target intents may lead to performance degradation. Toward this end, we propose a Multi-Behavior Intent Disentanglement (MBID) framework based on Information Bottleneck (IB) principle, which focuses on disentangling spurious correlation intents in multi-behavior recommendations. In particular, we design a projection-based intent extraction method to decompose the genuine and spurious correlation intents in auxiliary behaviors. Building on this, we conceive an IB-based multi-intent learning task to disentangle the spurious correlation intents and transfer the genuine correlation intents from auxiliary behaviors into the target behavior, yielding high-quality target intent representations. Experiments on three real-world datasets show MBID significantly outperforms the state-of-the-art baselines by effectively disentangling the spurious correlation intents.

Graph Convolutional Networks (GCNs) have become increasingly popular in recommendation systems. However, recent studies have shown that GCN-based models will cause sensitive information to disseminate widely in the graph structure, amplifying data bias and raising fairness concerns. While various fairness methods have been proposed, most of them neglect the impact of biased data on representation learning, which results in limited fairness improvement. Moreover, some studies have focused on constructing fair and balanced data distributions through data augmentation, but these methods significantly reduce utility due to disruption of user preferences. In this paper, we aim to design a fair recommendation method from the perspective of data augmentation to improve fairness while preserving recommendation utility. To achieve fairness-aware data augmentation with minimal disruption to user preferences, we propose two prior hypotheses. The first hypothesis identifies sensitive interactions by comparing outcomes of performance-oriented and fairness-aware recommendations, while the second one focuses on detecting sensitive features by analyzing feature similarities between biased and debiased representations. Then, we propose a dual data augmentation framework for fair recommendation, which includes two data augmentation strategies to generate fair augmented graphs and feature representations. Furthermore, we introduce a debiasing learning method that minimizes the dependence between the learned representations and sensitive information to eliminate bias. Extensive experiments on two real-world datasets demonstrate the superiority of our proposed framework.

Cihan Xiao, Matthew Wiesner, Debashish Chakraborty, Reno Kriz, Keith Cunningham, Kenton Murray, Kevin Duh, Luis Tavarez-Arce, Paul McNamee, Sanjeev Khudanpur. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Automatic speech recognition (ASR) systems can benefit from incorporating contextual information to improve recognition accuracy, especially for uncommon words or phrases. Current approaches like custom vocabularies or prompting with previous transcript segments provide limited contextual control. Compared to existing context biasing methods, RAG promises more flexible and scalable contextual control by leveraging LLMs’ broad knowledge. To this end, we propose leveraging large language models (LLMs) and retrieval-augmented generation (RAG) to enhance the contextual capabilities of ASR systems. Specifically, we propose systems based on text and audio LLMs to perform contextual error correction with context retrieved by querying a text-based retriever using the ASR module’s firstpass ASR hypotheses and a frequency-based custom vocabulary (CV) list. Our experiments reveal that the fine-tuned system has effectively learned to extract the relevant context to perform error correction while maintaining robustness against noise.