The increasing demand for personalized interactions with large language models (LLMs) calls for methodologies capable of accurately and efficiently identifying user opinions and preferences. Retrieval augmentation emerges as an effective strategy, as it can accommodate a vast number of users without the costs from fine-tuning. Existing research, however, has largely focused on enhancing the retrieval stage and devoted limited exploration toward optimizing the representation of the database, a crucial aspect for tasks such as personalization. In this work, we examine the problem from a novel angle, focusing on how data can be better represented for more data-efficient retrieval in the context of LLM customization. To tackle this challenge, we introduce Persona-DB, a simple yet effective framework consisting of a hierarchical construction process to improve generalization across task contexts and collaborative refinement to effectively bridge knowledge gaps among users. In the evaluation of response prediction, Persona-DB demonstrates superior context efficiency in maintaining accuracy with a significantly reduced retrieval size, a critical advantage in scenarios with extensive histories or limited context windows. Our experiments also indicate a marked improvement of over 10% under cold-start scenarios, when users have extremely sparse data. Furthermore, our analysis reveals the increasing importance of collaborative knowledge as the retrieval capacity expands.

The perceptions and decisions of individuals on social networks are deeply rooted in their intrinsic beliefs, which makes it possible to infer social beliefs from user behavior and message interactions. While existing research models these interactions as graphs and learns their representations, interpretability remains a significant challenge. In real-world scenarios, the interpretation of beliefs is nested within subject scopes of different granularity (such as topics and locations), posing additional challenges for belief discovery. In this paper, we introduce the Interpretable Graph Auto-Encoder Tree (IGAT), a novel end-to-end framework that jointly encodes hierarchical subject scopes and corresponding beliefs as a unified, interpretable hierarchical representation. IGAT integrates the interpretable hierarchy of Model Trees with disentangled representation learning models. We propose a differentiable Slice Mechanism to dynamically optimize internal node splitting and jointly train a leaf model to learn disentangled belief subspaces. The aggregation of these subspaces yields a unified representation, offering interpretations for both subjects and beliefs. Experimental evaluations on three real-world Twitter datasets show that IGAT achieves a consistent improvement of 1.49%-5.61% in F1-score, accuracy, and purity in the belief discovery task, as well as its effectiveness in various downstream analytical applications.

Chi Han, Jialiang Xu, Manling Li, Yi Fung, Chenkai Sun, Nan Jiang, Tarek Abdelzaher, Heng Ji. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

The paper advances belief representation learning in polarized networks – the mapping of social beliefs espoused by users and posts in a polarized network into a disentangled latent space that separates (the members and beliefs of) each side. Our prior work embeds social interaction data, using non-negative variational graph auto-encoders, into a disentangled latent space. However, the interaction graphs alone may not adequately reflect similarity and/or disparity in beliefs, especially for those graphs with sparsity and outlier issues. In this paper, we investigate the impact of limited guidance from Large Language Models (LLMs) on the accuracy of belief separation. Specifically, we integrate social graphs with LLM-based soft labels as a novel weakly-supervised interpretable graph representation learning framework. This framework combines the strengths of graph-and text-based information, and is shown to maintain the interpretability of learned representations, where different axes in the latent space denote association with different sides of the divide. An evaluation on six real-world Twitter datasets illustrates the effectiveness of the proposed model at solving stance detection problems, demonstrating 5.9%-6.5% improvements in the accuracy, F1 score, and purity metrics, without introducing a significant computational overhead. An ablation study is also discussed to study the impact of different components of the proposed architecture.

Introduced to enhance the efficiency of large language model (LLM) inference, speculative decoding operates by having a smaller model generate a draft. A larger target model then reviews this draft to align with its output, and any acceptance by the target model results in a reduction of the number of the target model runs, ultimately improving efficiency. However, the drafting process in speculative decoding includes slow autoregressive generation and allocates equal time to generating tokens, irrespective of their importance. These inefficiencies collectively contribute to the suboptimal performance of speculative decoding. To further improve LLM inference, we introduce Cascade Speculative Drafting (CS Drafting), a speculative execution algorithm that incorporates two types of cascades. The *Vertical Cascade* eliminates autoregressive generation from neural models, while the *Horizontal Cascade* optimizes time allocation in drafting for improved efficiency. Combining both cascades, CS Drafting achieves greater speedup compared to the baselines in our experiments, while preserving the same output distribution as the target model. Our code is publicly available at https://github.com/lfsszd/CS-Drafting.

In this paper, we present TETRIS, a new task of Goal-Oriented Script Completion.Unlike previous work, it considers a more realistic and general setting, where the input includes not only the goal but also additional user context, including preferences and history.To address this problem, we propose a novel approach, which uses two techniques to improve performance: (1) concept prompting, and (2) scriptoriented contrastive learning that addresses step repetition and hallucination problems.On our WikiHow-based dataset, we find that both methods improve performance.

Automatic response forecasting for news media plays a crucial role in enabling content producers to efficiently predict the impact of news releases and prevent unexpected negative outcomes such as social conflict and moral injury. To effectively forecast responses, it is essential to develop measures that leverage the social dynamics and contextual information surrounding individuals, especially in cases where explicit profiles or historical actions of the users are limited (referred to as lurkers). As shown in a previous study, 97% of all tweets are produced by only the most active 25% of users. However, existing approaches have limited exploration of how to best process and utilize these important features. To address this gap, we propose a novel framework, named SOCIALSENSE, that leverages a large language model to induce a belief-centered graph on top of an existent social network, along with graph-based propagation to capture social dynamics. We hypothesize that the induced graph that bridges the gap between distant users who share similar beliefs allows the model to effectively capture the response patterns. Our method surpasses existing state-of-the-art in experimental evaluations for both zero-shot and supervised settings, demonstrating its effectiveness in response forecasting. Moreover, the analysis reveals the framework's capability to effectively handle unseen user and lurker scenarios, further highlighting its robustness and practical applicability.

Automated knowledge discovery from trending chemical literature is essential for more efficient biomedical research. How to extract detailed knowledge about chemical reactions from the core chemistry literature is a new emerging challenge that has not been well studied. In this paper, we study the new problem of fine-grained chemical entity typing, which poses interesting new challenges especially because of the complex name mentions frequently occurring in chemistry literature and graphic representation of entities. We introduce a new benchmark data set (CHEMET) to facilitate the study of the new task and propose a novel multi-modal representation learning framework to solve the problem of fine-grained chemical entity typing by leveraging external resources with chemical structures and using cross-modal attention to learn effective representation of text in the chemistry domain. Experiment results show that the proposed framework outperforms multiple state-of-the-art methods.1

Data-driven technologies are only as good as the data they work with. On the other hand, data scientists have often limited control on how the data is collected. Failing to contain adequate number of instances from minority (sub)groups, known as population bias, is a major reason for model unfairness and disparate performance across different groups. We demonstrate MithraCoverage, a system for investigating population bias over the intersection of multiple attributes. We use the concept of coverage for identifying intersectional subgroups with inadequate representation in the dataset. MithraCoverage is a web application with an interactive visual interface that allows data scientists to explore the dataset and identify subgroups with poor coverage.

Using inappropriate datasets for data science tasks can be harmful, especially for applications that impact humans. Targeting data ethics, we demonstrate MithraLabel, a system for generating task-specific information about a dataset, in the form of a set of visual widgets, as a flexible "nutritional label" that provides a user with information to determine the fitness of the dataset for the task at hand.