Dongha Lee

Su Kim, Dongha Lee, SeongKu Kang et al.

Recently, graph neural networks (GNNs) have been successfully applied to predicting molecular properties, which is one of the most classical cheminformatics tasks with various applications. Despite their effectiveness, we empirically observe that training a single GNN model for diverse molecules with distinct structural patterns limits its prediction performance. In this paper, motivated by this observation, we propose TopExpert to leverage topology-specific prediction models (referred to as experts), each of which is responsible for each molecular group sharing similar topological semantics. That is, each expert learns topology-specific discriminative features while being trained with its corresponding topological group. To tackle the key challenge of grouping molecules by their topological patterns, we introduce a clustering-based gating module that assigns an input molecule into one of the clusters and further optimizes the gating module with two different types of self-supervision: topological semantics induced by GNNs and molecular scaffolds, respectively. Extensive experiments demonstrate that TopExpert has boosted the performance for molecular property prediction and also achieved better generalization for new molecules with unseen scaffolds than baselines. The code is available at https://github.com/kimsu55/ToxExpert.

Dongha Lee, Jiaming Shen, Seonghyeon Lee et al. · deepmind

Topic taxonomies display hierarchical topic structures of a text corpus and provide topical knowledge to enhance various NLP applications. To dynamically incorporate new topic information, several recent studies have tried to expand (or complete) a topic taxonomy by inserting emerging topics identified in a set of new documents. However, existing methods focus only on frequent terms in documents and the local topic-subtopic relations in a taxonomy, which leads to limited topic term coverage and fails to model the global topic hierarchy. In this work, we propose a novel framework for topic taxonomy expansion, named TopicExpan, which directly generates topic-related terms belonging to new topics. Specifically, TopicExpan leverages the hierarchical relation structure surrounding a new topic and the textual content of an input document for topic term generation. This approach encourages newly-inserted topics to further cover important but less frequent terms as well as to keep their relation consistency within the taxonomy. Experimental results on two real-world text corpora show that TopicExpan significantly outperforms other baseline methods in terms of the quality of output taxonomies.

Hyungjoo Chae, Yongho Song, Kai Tzu-iunn Ong et al. · gatech

Human-like chatbots necessitate the use of commonsense reasoning in order to effectively comprehend and respond to implicit information present within conversations. Achieving such coherence and informativeness in responses, however, is a non-trivial task. Even for large language models (LLMs), the task of identifying and aggregating key evidence within a single hop presents a substantial challenge. This complexity arises because such evidence is scattered across multiple turns in a conversation, thus necessitating integration over multiple hops. Hence, our focus is to facilitate such multi-hop reasoning over a dialogue context, namely dialogue chain-of-thought (CoT) reasoning. To this end, we propose a knowledge distillation framework that leverages LLMs as unreliable teachers and selectively distills consistent and helpful rationales via alignment filters. We further present DOCTOR, a DialOgue Chain-of-ThOught Reasoner that provides reliable CoT rationales for response generation. We conduct extensive experiments to show that enhancing dialogue agents with high-quality rationales from DOCTOR significantly improves the quality of their responses.

Susik Yoon, Yu Meng, Dongha Lee et al.

We present a framework SCStory for online story discovery, that helps people digest rapidly published news article streams in real-time without human annotations. To organize news article streams into stories, existing approaches directly encode the articles and cluster them based on representation similarity. However, these methods yield noisy and inaccurate story discovery results because the generic article embeddings do not effectively reflect the story-indicative semantics in an article and cannot adapt to the rapidly evolving news article streams. SCStory employs self-supervised and continual learning with a novel idea of story-indicative adaptive modeling of news article streams. With a lightweight hierarchical embedding module that first learns sentence representations and then article representations, SCStory identifies story-relevant information of news articles and uses them to discover stories. The embedding module is continuously updated to adapt to evolving news streams with a contrastive learning objective, backed up by two unique techniques, confidence-aware memory replay and prioritized-augmentation, employed for label absence and data scarcity problems. Thorough experiments on real and the latest news data sets demonstrate that SCStory outperforms existing state-of-the-art algorithms for unsupervised online story discovery.

Susik Yoon, Dongha Lee, Yunyi Zhang et al. · amazon-science

Unsupervised discovery of stories with correlated news articles in real-time helps people digest massive news streams without expensive human annotations. A common approach of the existing studies for unsupervised online story discovery is to represent news articles with symbolic- or graph-based embedding and incrementally cluster them into stories. Recent large language models are expected to improve the embedding further, but a straightforward adoption of the models by indiscriminately encoding all information in articles is ineffective to deal with text-rich and evolving news streams. In this work, we propose a novel thematic embedding with an off-the-shelf pretrained sentence encoder to dynamically represent articles and stories by considering their shared temporal themes. To realize the idea for unsupervised online story discovery, a scalable framework USTORY is introduced with two main techniques, theme- and time-aware dynamic embedding and novelty-aware adaptive clustering, fueled by lightweight story summaries. A thorough evaluation with real news data sets demonstrates that USTORY achieves higher story discovery performances than baselines while being robust and scalable to various streaming settings.

SeongKu Kang, Wonbin Kweon, Dongha Lee et al.

Recent recommender systems have shown remarkable performance by using an ensemble of heterogeneous models. However, it is exceedingly costly because it requires resources and inference latency proportional to the number of models, which remains the bottleneck for production. Our work aims to transfer the ensemble knowledge of heterogeneous teachers to a lightweight student model using knowledge distillation (KD), to reduce the huge inference costs while retaining high accuracy. Through an empirical study, we find that the efficacy of distillation severely drops when transferring knowledge from heterogeneous teachers. Nevertheless, we show that an important signal to ease the difficulty can be obtained from the teacher's training trajectory. This paper proposes a new KD framework, named HetComp, that guides the student model by transferring easy-to-hard sequences of knowledge generated from the teachers' trajectories. To provide guidance according to the student's learning state, HetComp uses dynamic knowledge construction to provide progressively difficult ranking knowledge and adaptive knowledge transfer to gradually transfer finer-grained ranking information. Our comprehensive experiments show that HetComp significantly improves the distillation quality and the generalization of the student model.

Kwangwook Seo, Dongha Lee

Recent approaches in personalized reward modeling have primarily focused on leveraging user interaction history to align model judgments with individual preferences. However, existing approaches largely treat user context as a static or implicit conditioning signal, failing to capture the dynamic and multi-faceted nature of human judgment. In this paper, we propose P-Check, a novel personalized reward modeling framework, designed to train a plug-and-play checklist generator that synthesizes dynamic evaluation criteria for guiding the reward prediction. To better align these checklists with personalized nuances, we introduce Preference-Contrastive Criterion Weighting, a training strategy that assigns saliency scores to criteria based on their discriminative power for personalized judgment. We conduct extensive experiments and demonstrate that P-Check not only improves reward accuracy but also enhances downstream personalized generation, and remains robust in OOD scenarios.

Seonghyeon Lee, Suyeon Kim, Joonwon Jang et al.

We study the code generation behavior of instruction-tuned models built on top of code pre-trained language models when they could access an auxiliary function to implement a function. We design several ways to provide auxiliary functions to the models by adding them to the query or providing a response prefix to incorporate the ability to utilize auxiliary functions with the instruction-following capability. Our experimental results show the effectiveness of combining the base models' auxiliary function utilization ability with the instruction following ability. In particular, the performance of adopting our approaches with the open-sourced language models surpasses that of the recent powerful proprietary language models, i.e., gpt-4o.

Jieyong Kim, Hyunseo Kim, Hyunjin Cho et al.

Recent advancements in Large Language Models (LLMs) have demonstrated exceptional performance across a wide range of tasks, generating significant interest in their application to recommendation systems. However, existing methods have not fully capitalized on the potential of LLMs, often constrained by limited input information or failing to fully utilize their advanced reasoning capabilities. To address these limitations, we introduce EXP3RT, a novel LLM-based recommender designed to leverage rich preference information contained in user and item reviews. EXP3RT is basically fine-tuned through distillation from a teacher LLM to perform three key tasks in order: EXP3RT first extracts and encapsulates essential subjective preferences from raw reviews, aggregates and summarizes them according to specific criteria to create user and item profiles. It then generates detailed step-by-step reasoning followed by predicted rating, i.e., reasoning-enhanced rating prediction, by considering both subjective and objective information from user/item profiles and item descriptions. This personalized preference reasoning from EXP3RT enhances rating prediction accuracy and also provides faithful and reasonable explanations for recommendation. Extensive experiments show that EXP3RT outperforms existing methods on both rating prediction and candidate item reranking for top-k recommendation, while significantly enhancing the explainability of recommendation systems.

Soojin Yoon, Sungho Ko, Tongyoung Kim et al.

Cross-lingual entity alignment (EA) enables the integration of multiple knowledge graphs (KGs) across different languages, providing users with seamless access to diverse and comprehensive knowledge. Existing methods, mostly supervised, face challenges in obtaining labeled entity pairs. To address this, recent studies have shifted towards self-supervised and unsupervised frameworks. Despite their effectiveness, these approaches have limitations: (1) Relation passing: mainly focusing on the entity while neglecting the semantic information of relations, (2) Isomorphic assumption: assuming isomorphism between source and target graphs, which leads to noise and reduced alignment accuracy, and (3) Noise vulnerability: susceptible to noise in the textual features, especially when encountering inconsistent translations or Out-of-Vocabulary (OOV) problems. In this paper, we propose ERAlign, an unsupervised and robust cross-lingual EA pipeline that jointly performs Entity-level and Relation-level Alignment by neighbor triple matching strategy using semantic textual features of relations and entities. Its refinement step iteratively enhances results by fusing entity-level and relation-level alignments based on neighbor triple matching. The additional verification step examines the entities' neighbor triples as the linearized text. This Align-then-Verify pipeline rigorously assesses alignment results, achieving near-perfect alignment even in the presence of noisy textual features of entities. Our extensive experiments demonstrate that the robustness and general applicability of ERAlign improved the accuracy and effectiveness of EA tasks, contributing significantly to knowledge-oriented applications.

Heejae Chon, Seonghyeon Lee, Jinyoung Yeo et al.

Language models (LMs) have exhibited impressive abilities in generating codes from natural language requirements. In this work, we highlight the diversity of code generated by LMs as a critical criterion for evaluating their code generation capabilities, in addition to functional correctness. Despite its practical implications, there is a lack of studies focused on assessing the diversity of generated code, which overlooks its importance in the development of code LMs. We propose a systematic approach to evaluate the diversity of generated code, utilizing various metrics for inter-code similarity as well as functional correctness. Specifically, we introduce a pairwise code similarity measure that leverages large LMs' capabilities in code understanding and reasoning, demonstrating the highest correlation with human judgment. We extensively investigate the impact of various factors on the quality of generated code, including model sizes, temperatures, training approaches, prompting strategies, and the difficulty of input problems. Our consistent observation of a positive correlation between the test pass score and the inter-code similarity score indicates that current LMs tend to produce functionally correct code with limited diversity.

Dongil Yang, Suyeon Lee, Minjin Kim et al.

Engagement between instructors and students plays a crucial role in enhancing students'academic performance. However, instructors often struggle to provide timely and personalized support in large classes. To address this challenge, we propose a novel Virtual Teaching Assistant (VTA) named YA-TA, designed to offer responses to students that are grounded in lectures and are easy to understand. To facilitate YA-TA, we introduce the Dual Retrieval-augmented Knowledge Fusion (DRAKE) framework, which incorporates dual retrieval of instructor and student knowledge and knowledge fusion for tailored response generation. Experiments conducted in real-world classroom settings demonstrate that the DRAKE framework excels in aligning responses with knowledge retrieved from both instructor and student sides. Furthermore, we offer additional extensions of YA-TA, such as a Q&A board and self-practice tools to enhance the overall learning experience. Our video is publicly available.

Seonglae Cho, Yonggi Cho, HoonJae Lee et al.

In this paper, we present RTSUM, an unsupervised summarization framework that utilizes relation triples as the basic unit for summarization. Given an input document, RTSUM first selects salient relation triples via multi-level salience scoring and then generates a concise summary from the selected relation triples by using a text-to-text language model. On the basis of RTSUM, we also develop a web demo for an interpretable summarizing tool, providing fine-grained interpretations with the output summary. With support for customization options, our tool visualizes the salience for textual units at three distinct levels: sentences, relation triples, and phrases. The codes,are publicly available.

Jeongeun Lee, Seongku Kang, Won-Yong Shin et al.

Collaborative Filtering (CF) is a foundational approach in recommender systems, but it struggles with challenges such as data sparsity and the cold-start problem. Cross-Domain Recommendation (CDR) has emerged as a promising solution by leveraging dense domains to improve recommendations in sparse target domains. However, existing CDR methods face significant limitations, including their reliance on overlapping users as a bridge between domains and their inability to address domain sensitivity, i.e., differences in user behaviors and characteristics across domains, effectively. To overcome these limitations, we propose CGSP, a unified and adaptive CDR framework based on graph signal processing (GSP). CGSP supports both intra-domain and inter-domain recommendations while adaptively controlling the influence of the source domain through a simple hyperparameter. The framework constructs a cross-domain similarity graph by integrating target-only and source-bridged similarity graphs to capture both intra-domain and inter-domain relationships. This graph is then processed through graph filtering techniques to propagate and enhance local signals. Finally, personalized graph signals are constructed, tailored separately for users in the source and target domains, enabling CGSP to function as a unified framework for CDR scenarios. Extensive evaluation shows that CGSP outperforms state-of-the-art baselines across diverse cross-domain settings, with notable gains in low-overlap scenarios, underscoring its practicality for real-world applications.

Sungho Ko, Hyunjin Cho, Hyungjoo Chae et al. · gatech

Recent studies have investigated utilizing Knowledge Graphs (KGs) to enhance Quesetion Answering (QA) performance of Large Language Models (LLMs), yet structured KG verbalization remains challengin. Existing methods, such as triple-form or free-form textual conversion of triple-form facts, encounter several issues. These include reduced evidence density due to duplicated entities or relationships, and reduced evidence clarity due to an inability to emphasize crucial evidence. To address these issues, we propose EFSum, an Evidence-focused Fact Summarization framework for enhanced QA with knowledge-augmented LLMs. We optimize an open-source LLM as a fact summarizer through distillation and preference alignment. Our extensive experiments show that EFSum improves LLM's zero-shot QA performance, and it is possible to ensure both the helpfulness and faithfulness of the summary.

Seonghyeon Lee, Sanghwan Jang, Seongbo Jang et al.

Auxiliary function is a helpful component to improve language model's code generation ability. However, a systematic exploration of how they affect has yet to be done. In this work, we comprehensively evaluate the ability to utilize auxiliary functions encoded in recent code-pretrained language models. First, we construct a human-crafted evaluation set, called HumanExtension, which contains examples of two functions where one function assists the other. With HumanExtension, we design several experiments to examine their ability in a multifaceted way. Our evaluation processes enable a comprehensive understanding of including auxiliary functions in the prompt in terms of effectiveness and robustness. An additional implementation style analysis captures the models' various implementation patterns when they access the auxiliary function. Through this analysis, we discover the models' promising ability to utilize auxiliary functions including their self-improving behavior by implementing the two functions step-by-step. However, our analysis also reveals the model's underutilized behavior to call the auxiliary function, suggesting the future direction to enhance their implementation by eliciting the auxiliary function call ability encoded in the models. We release our code and dataset to facilitate this research direction.

Seonghyeon Lee, Heejae Chon, Joonwon Jang et al.

Language models (LMs) have exhibited impressive abilities in generating code from natural language requirements. In this work, we highlight the diversity of code generated by LMs as a critical criterion for evaluating their code generation capabilities. There is a lack of studies focused on assessing the diversity of generated code, which overlooks its importance in code LMs. Therefore, we propose a systematic approach to evaluate code diversity, introducing various metrics with inter-code similarity. Specifically, we introduce code clustering methods that leverages LMs' capabilities in code understanding and reasoning, resulting in a set of metrics that represent the number of algorithms in model-generated solutions. We extensively investigate the property of model-generated solutions by contrasting them with human-written ones and quantifying the impact of various factors on code diversity: model size, temperature, instruction tuning, and problem complexity. Our analysis demonstrates that model-generated solutions exhibit low algorithmic diversity, which was neglected by the research community. Moreover, we explore methods to increase code diversity by combining solutions from different models and increasing sampling temperatures. Our findings highlight that code diversity can be enhanced with the help of heterogeneous models and setting temperature beyond 1.0 that has not been fully explored due to the functional correctness degradation. To facilitate our research direction, we publicly share our code and datasets through open-source repositories.

SeongKu Kang, Jianxun Lian, Dongha Lee et al.

Recommender systems suffer from biases that cause the collected feedback to incompletely reveal user preference. While debiasing learning has been extensively studied, they mostly focused on the specialized (called counterfactual) test environment simulated by random exposure of items, significantly degrading accuracy in the typical (called factual) test environment based on actual user-item interactions. In fact, each test environment highlights the benefit of a different aspect: the counterfactual test emphasizes user satisfaction in the long-terms, while the factual test focuses on predicting subsequent user behaviors on platforms. Therefore, it is desirable to have a model that performs well on both tests rather than only one. In this work, we introduce a new learning framework, called Bias-adaptive Preference distillation Learning (BPL), to gradually uncover user preferences with dual distillation strategies. These distillation strategies are designed to drive high performance in both factual and counterfactual test environments. Employing a specialized form of teacher-student distillation from a biased model, BPL retains accurate preference knowledge aligned with the collected feedback, leading to high performance in the factual test. Furthermore, through self-distillation with reliability filtering, BPL iteratively refines its knowledge throughout the training process. This enables the model to produce more accurate predictions across a broader range of user-item combinations, thereby improving performance in the counterfactual test. Comprehensive experiments validate the effectiveness of BPL in both factual and counterfactual tests. Our implementation is accessible via: https://github.com/SeongKu-Kang/BPL.

Wonjoong Kim, Sangwu Park, Yeonjun In et al.

Although recent tool-augmented benchmarks incorporate complex user requests and diverse tools, the evaluation methods for most of them remain limited to answer matching. However, as the number of steps required to resolve a user request increases, a proper evaluation of an agent's performance must go beyond the final answer to also assess the problem-solving trajectory, including previously ignored aspects such as efficiency, hallucination, and adaptivity. The most straightforward method for evaluating these aspects is to compare an agent's trajectory with the ground-truth trajectory, but this approach is fundamentally limited since annotating all valid ground-truth trajectories is prohibitively expensive. However, a simple LLM-based evaluator struggles to assess trajectories in detail without ground truth. To effectively evaluate the agents in this manner, we introduce TRACE, a framework for the multi-dimensional evaluation of tool-augmented LLM agent performance. By incorporating an evidence bank, which accumulates knowledge gathered from preceding reasoning steps, TRACE enables a multi-faceted analysis and evaluation of an agent's reasoning trajectory effectively. To validate our framework, we develop a new meta-evaluation dataset by augmenting existing benchmarks with diverse and flawed trajectories, each labeled with multi-faceted performance scores. Our results confirm that TRACE accurately evaluates these complex behaviors in a scalable and cost-effective manner, even with small open-source LLMs. Furthermore, we apply our method to evaluate the trajectories that agents produce while solving tool-augmented tasks, presenting previously unreported observations and their corresponding insights.

Hyunseo Kim, Sangam Lee, Kwangwook Seo et al.

Search-augmented large language models (LLMs) have advanced information-seeking tasks by integrating retrieval into generation, reducing users' cognitive burden compared to traditional search systems. Yet they remain insufficient for fully addressing diverse user needs, which requires recognizing how the same query can reflect different intents across users and delivering information in preferred forms. While recent systems such as ChatGPT and Gemini attempt personalization by leveraging user histories, systematic evaluation of such personalization is under-explored. To address this gap, we propose BESPOKE, the realistic benchmark for evaluating personalization in search-augmented LLMs. BESPOKE is designed to be both realistic, by collecting authentic chat and search histories directly from humans, and diagnostic, by pairing responses with fine-grained preference scores and feedback. The benchmark is constructed through long-term, deeply engaged human annotation, where human annotators contributed their own histories, authored queries with detailed information needs, and evaluated responses with scores and diagnostic feedback. Leveraging BESPOKE, we conduct systematic analyses that reveal key requirements for effective personalization in information-seeking tasks, providing a foundation for fine-grained evaluation of personalized search-augmented LLMs. Our code and data are available at https://augustinlib.github.io/BESPOKE/.

Kai Tzu-iunn Ong, Namyoung Kim, Minju Gwak et al.

To achieve lifelong human-agent interaction, dialogue agents need to constantly memorize perceived information and properly retrieve it for response generation (RG). While prior studies focus on getting rid of outdated memories to improve retrieval quality, we argue that such memories provide rich, important contextual cues for RG (e.g., changes in user behaviors) in long-term conversations. We present THEANINE, a framework for LLM-based lifelong dialogue agents. THEANINE discards memory removal and manages large-scale memories by linking them based on their temporal and cause-effect relation. Enabled by this linking structure, THEANINE augments RG with memory timelines - series of memories representing the evolution or causality of relevant past events. Along with THEANINE, we introduce TeaFarm, a counterfactual-driven evaluation scheme, addressing the limitation of G-Eval and human efforts when assessing agent performance in integrating past memories into RG. A supplementary video for THEANINE and data for TeaFarm are at https://huggingface.co/spaces/ResearcherScholar/Theanine.

Seonghyeon Lee, Dongha Lee, Seongbo Jang et al.

Recently, finetuning a pretrained language model to capture the similarity between sentence embeddings has shown the state-of-the-art performance on the semantic textual similarity (STS) task. However, the absence of an interpretation method for the sentence similarity makes it difficult to explain the model output. In this work, we explicitly describe the sentence distance as the weighted sum of contextualized token distances on the basis of a transportation problem, and then present the optimal transport-based distance measure, named RCMD; it identifies and leverages semantically-aligned token pairs. In the end, we propose CLRCMD, a contrastive learning framework that optimizes RCMD of sentence pairs, which enhances the quality of sentence similarity and their interpretation. Extensive experiments demonstrate that our learning framework outperforms other baselines on both STS and interpretable-STS benchmarks, indicating that it computes effective sentence similarity and also provides interpretation consistent with human judgement. The code and checkpoint are publicly available at https://github.com/sh0416/clrcmd.

Hyungjoo Chae, Namyoung Kim, Kai Tzu-iunn Ong et al. · gatech

Large language models (LLMs) have recently gained much attention in building autonomous agents. However, the performance of current LLM-based web agents in long-horizon tasks is far from optimal, often yielding errors such as repeatedly buying a non-refundable flight ticket. By contrast, humans can avoid such an irreversible mistake, as we have an awareness of the potential outcomes (e.g., losing money) of our actions, also known as the "world model". Motivated by this, our study first starts with preliminary analyses, confirming the absence of world models in current LLMs (e.g., GPT-4o, Claude-3.5-Sonnet, etc.). Then, we present a World-model-augmented (WMA) web agent, which simulates the outcomes of its actions for better decision-making. To overcome the challenges in training LLMs as world models predicting next observations, such as repeated elements across observations and long HTML inputs, we propose a transition-focused observation abstraction, where the prediction objectives are free-form natural language descriptions exclusively highlighting important state differences between time steps. Experiments on WebArena and Mind2Web show that our world models improve agents' policy selection without training and demonstrate our agents' cost- and time-efficiency compared to recent tree-search-based agents.

Minjin Kim, Minju Kim, Hana Kim et al.

Conversational recommender system is an emerging area that has garnered an increasing interest in the community, especially with the advancements in large language models (LLMs) that enable diverse reasoning over conversational input. Despite the progress, the field has many aspects left to explore. The currently available public datasets for conversational recommendation lack specific user preferences and explanations for recommendations, hindering high-quality recommendations. To address such challenges, we present a novel conversational recommendation dataset named PEARL, synthesized with persona- and knowledge-augmented LLM simulators. We obtain detailed persona and knowledge from real-world reviews and construct a large-scale dataset with over 57k dialogues. Our experimental results demonstrate that utterances in PEARL include more specific user preferences, show expertise in the target domain, and provide recommendations more relevant to the dialogue context than those in prior datasets.

SeongKu Kang, Shivam Agarwal, Bowen Jin et al.

Document retrieval has greatly benefited from the advancements of large-scale pre-trained language models (PLMs). However, their effectiveness is often limited in theme-specific applications for specialized areas or industries, due to unique terminologies, incomplete contexts of user queries, and specialized search intents. To capture the theme-specific information and improve retrieval, we propose to use a corpus topical taxonomy, which outlines the latent topic structure of the corpus while reflecting user-interested aspects. We introduce ToTER (Topical Taxonomy Enhanced Retrieval) framework, which identifies the central topics of queries and documents with the guidance of the taxonomy, and exploits their topical relatedness to supplement missing contexts. As a plug-and-play framework, ToTER can be flexibly employed to enhance various PLM-based retrievers. Through extensive quantitative, ablative, and exploratory experiments on two real-world datasets, we ascertain the benefits of using topical taxonomy for retrieval in theme-specific applications and demonstrate the effectiveness of ToTER.

Jieyong Kim, Ryang Heo, Yongsik Seo et al.

In the task of aspect sentiment quad prediction (ASQP), generative methods for predicting sentiment quads have shown promising results. However, they still suffer from imprecise predictions and limited interpretability, caused by data scarcity and inadequate modeling of the quadruplet composition process. In this paper, we propose Self-Consistent Reasoning-based Aspect-sentiment quadruple Prediction (SCRAP), optimizing its model to generate reasonings and the corresponding sentiment quadruplets in sequence. SCRAP adopts the Extract-Then-Assign reasoning strategy, which closely mimics human cognition. In the end, SCRAP significantly improves the model's ability to handle complex reasoning tasks and correctly predict quadruplets through consistency voting, resulting in enhanced interpretability and accuracy in ASQP.

SeongKu Kang, Yunyi Zhang, Pengcheng Jiang et al. · amazon-science

Academic paper search is an essential task for efficient literature discovery and scientific advancement. While dense retrieval has advanced various ad-hoc searches, it often struggles to match the underlying academic concepts between queries and documents, which is critical for paper search. To enable effective academic concept matching for paper search, we propose Taxonomy-guided Semantic Indexing (TaxoIndex) framework. TaxoIndex extracts key concepts from papers and organizes them as a semantic index guided by an academic taxonomy, and then leverages this index as foundational knowledge to identify academic concepts and link queries and documents. As a plug-and-play framework, TaxoIndex can be flexibly employed to enhance existing dense retrievers. Extensive experiments show that TaxoIndex brings significant improvements, even with highly limited training data, and greatly enhances interpretability.

SeongKu Kang, Bowen Jin, Wonbin Kweon et al.

In specialized fields like the scientific domain, constructing large-scale human-annotated datasets poses a significant challenge due to the need for domain expertise. Recent methods have employed large language models to generate synthetic queries, which serve as proxies for actual user queries. However, they lack control over the content generated, often resulting in incomplete coverage of academic concepts in documents. We introduce Concept Coverage-based Query set Generation (CCQGen) framework, designed to generate a set of queries with comprehensive coverage of the document's concepts. A key distinction of CCQGen is that it adaptively adjusts the generation process based on the previously generated queries. We identify concepts not sufficiently covered by previous queries, and leverage them as conditions for subsequent query generation. This approach guides each new query to complement the previous ones, aiding in a thorough understanding of the document. Extensive experiments demonstrate that CCQGen significantly enhances query quality and retrieval performance.

Kwangwook Seo, Donguk Kwon, Dongha Lee

Recent advancements in table-based reasoning have expanded beyond factoid-level QA to address insight-level tasks, where systems should synthesize implicit knowledge in the table to provide explainable analyses. Although effective, existing studies remain confined to scenarios where a single gold table is given alongside the user query, failing to address cases where users seek comprehensive insights from multiple unknown tables. To bridge these gaps, we propose MT-RAIG Bench, design to evaluate systems on Retrieval-Augmented Insight Generation over Mulitple-Tables. Additionally, to tackle the suboptimality of existing automatic evaluation methods in the table domain, we further introduce a fine-grained evaluation framework MT-RAIG Eval, which achieves better alignment with human quality judgments on the generated insights. We conduct extensive experiments and reveal that even frontier LLMs still struggle with complex multi-table reasoning, establishing our MT-RAIG Bench as a challenging testbed for future research.

Seoyeon Kim, Kwangwook Seo, Hyungjoo Chae et al. · gatech

Recent approaches in domain-specific named entity recognition (NER), such as biomedical NER, have shown remarkable advances. However, they still lack of faithfulness, producing erroneous predictions. We assume that knowledge of entities can be useful in verifying the correctness of the predictions. Despite the usefulness of knowledge, resolving such errors with knowledge is nontrivial, since the knowledge itself does not directly indicate the ground-truth label. To this end, we propose VerifiNER, a post-hoc verification framework that identifies errors from existing NER methods using knowledge and revises them into more faithful predictions. Our framework leverages the reasoning abilities of large language models to adequately ground on knowledge and the contextual information in the verification process. We validate effectiveness of VerifiNER through extensive experiments on biomedical datasets. The results suggest that VerifiNER can successfully verify errors from existing models as a model-agnostic approach. Further analyses on out-of-domain and low-resource settings show the usefulness of VerifiNER on real-world applications.

Seoyeon Kim, Huiseo Kim, Chanjun Park et al.

Recent large language models (LLMs) demonstrate multilingual abilities, yet they are English-centric due to dominance of English in training corpora. The limited resource for low-resource languages remains a crucial challenge. Code-switching (CS), a phenomenon where multilingual speakers alternate between languages in a discourse, can convey subtle cultural and linguistic nuances that can be otherwise lost in translation and elicits language-specific knowledge in human communications. In light of this, we investigate whether code-switching can activate, or identify and leverage knowledge for reasoning when LLMs solve low-resource language tasks. To facilitate the research, we first present EnKoQA, a synthetic English-Korean CS question-answering dataset. We provide comprehensive analysis on a variety of multilingual LLMs by subdividing activation process into knowledge identification and knowledge leveraging. Our results demonstrate that compared to English text, CS can faithfully activate knowledge inside LLMs especially on language-specific domains, suggesting the potential of code-switching on low-resource language tasks.

Jieyong Kim, Tongyoung Kim, Soojin Yoon et al.

While black-box large language models are widely deployed, they produce generic outputs that overlook individual user preferences. Current personalization methods are fundamentally limited to response-level personalization; they only match final outputs, failing to model the underlying reasoning that connects user behavior to responses. To address this, this work introduces reasoning-level personalization as a new paradigm and proposes RPM, the first systematic framework designed to guide the model's reasoning process using structured rationales constructed from patterns in a user's behavior. RPM constructs a structured model of user behavior-built from response-influential features and statistical factors-to create personalized reasoning paths and retrieve beneficial examples for guiding inference through a feature-based retrieval mechanism. Extensive experiments across four diverse tasks demonstrate that RPM consistently outperforms existing response-level methods while simultaneously enhancing both personalization performance and interpretability, providing a promising direction for black-box LLM personalization.

Sunghwan Kim, Dongjin Kang, Taeyoon Kwon et al. · gatech

Reward models (RMs) play a crucial role in reinforcement learning from human feedback (RLHF), aligning model behavior with human preferences. However, existing benchmarks for reward models show a weak correlation with the performance of optimized policies, suggesting that they fail to accurately assess the true capabilities of RMs. To bridge this gap, we explore several evaluation designs through the lens of reward overoptimization\textemdash a phenomenon that captures both how well the reward model aligns with human preferences and the dynamics of the learning signal it provides to the policy. The results highlight three key findings on how to construct a reliable benchmark: (i) it is important to minimize differences between chosen and rejected responses beyond correctness, (ii) evaluating reward models requires multiple comparisons across a wide range of chosen and rejected responses, and (iii) given that reward models encounter responses with diverse representations, responses should be sourced from a variety of models. However, we also observe that a extremely high correlation with degree of overoptimization leads to comparatively lower correlation with certain downstream performance. Thus, when designing a benchmark, it is desirable to use the degree of overoptimization as a useful tool, rather than the end goal.

Seo Hyun Kim, Keummin Ka, Yohan Jo et al.

For a human-like chatbot, constructing a long-term memory is crucial. However, current large language models often lack this capability, leading to instances of missing important user information or redundantly asking for the same information, thereby diminishing conversation quality. To effectively construct memory, it is crucial to seamlessly connect past and present information, while also possessing the ability to forget obstructive information. To address these challenges, we propose CREEM, a novel memory system for long-term conversation. Improving upon existing approaches that construct memory based solely on current sessions, CREEM blends past memories during memory formation. Additionally, we introduce a refining process to handle redundant or outdated information. Unlike traditional paradigms, we view responding and memory construction as inseparable tasks. The blending process, which creates new memories, also serves as a reasoning step for response generation by informing the connection between past and present. Through evaluation, we demonstrate that CREEM enhances both memory and response qualities in multi-session personalized dialogues.

Jeongeun Lee, Youngjae Yu, Dongha Lee

The exponential growth of video content has made personalized video highlighting an essential task, as user preferences are highly variable and complex. Existing video datasets, however, often lack personalization, relying on isolated videos or simple text queries that fail to capture the intricacies of user behavior. In this work, we introduce HIPPO-Video, a novel dataset for personalized video highlighting, created using an LLM-based user simulator to generate realistic watch histories reflecting diverse user preferences. The dataset includes 2,040 (watch history, saliency score) pairs, covering 20,400 videos across 170 semantic categories. To validate our dataset, we propose HiPHer, a method that leverages these personalized watch histories to predict preference-conditioned segment-wise saliency scores. Through extensive experiments, we demonstrate that our method outperforms existing generic and query-based approaches, showcasing its potential for highly user-centric video highlighting in real-world scenarios.

Seongbo Jang, Minjin Jeon, Jaehoon Lee et al.

While research on dialogue response generation has primarily focused on generating coherent responses conditioning on textual context, the critical question of when to respond grounded on the temporal context remains underexplored. To bridge this gap, we propose a novel task called timely dialogue response generation and introduce the TimelyChat benchmark, which evaluates the capabilities of language models to predict appropriate time intervals and generate time-conditioned responses. Additionally, we construct a large-scale training dataset by leveraging unlabeled event knowledge from a temporal commonsense knowledge graph and employing a large language model (LLM) to synthesize 55K event-driven dialogues. We then train Timer, a dialogue agent designed to proactively predict time intervals and generate timely responses that align with those intervals. Experimental results show that Timer outperforms prompting-based LLMs and other fine-tuned baselines in both turn-level and dialogue-level evaluations. We publicly release our data, model, and code.

Seongbo Jang, Seonghyeon Lee, Dongha Lee et al.

Multimodal chatbots have become one of the major topics for dialogue systems in both research community and industry. Recently, researchers have shed light on the multimodality of responses as well as dialogue contexts. This work explores how a dialogue system can output responses in various modalities such as text and image. To this end, we first formulate a multimodal dialogue response retrieval task for retrieval-based systems as the combination of three subtasks. We then propose three integration methods based on a two-step approach and an end-to-end approach, and compare the merits and demerits of each method. Experimental results on two datasets demonstrate that the end-to-end approach achieves comparable performance without an intermediate step in the two-step approach. In addition, a parameter sharing strategy not only reduces the number of parameters but also boosts performance by transferring knowledge across the subtasks and the modalities.

Ryang Heo, Yongsik Seo, Junseong Lee et al.

The surge of user-generated online content presents a wealth of insights into customer preferences and market trends. However, the highly diverse, complex, and context-rich nature of such contents poses significant challenges to traditional opinion mining approaches. To address this, we introduce Online Opinion Mining Benchmark (OOMB), a novel dataset and evaluation protocol designed to assess the ability of large language models (LLMs) to mine opinions effectively from diverse and intricate online environments. OOMB provides extensive (entity, feature, opinion) tuple annotations and a comprehensive opinion-centric summary that highlights key opinion topics within each content, thereby enabling the evaluation of both the extractive and abstractive capabilities of models. Through our proposed benchmark, we conduct a comprehensive analysis of which aspects remain challenging and where LLMs exhibit adaptability, to explore whether they can effectively serve as opinion miners in realistic online scenarios. This study lays the foundation for LLM-based opinion mining and discusses directions for future research in this field.

Kwangwook Seo, Jinyoung Yeo, Dongha Lee

Implicit knowledge hidden within the explicit table cells, such as data insights, is the key to generating a high-quality table summary. However, unveiling such implicit knowledge is a non-trivial task. Due to the complex nature of structured tables, it is challenging even for large language models (LLMs) to mine the implicit knowledge in an insightful and faithful manner. To address this challenge, we propose a novel table reasoning framework Question-then-Pinpoint. Our work focuses on building a plug-and-play table reasoner that can self-question the insightful knowledge and answer it by faithfully pinpointing evidence on the table to provide explainable guidance for the summarizer. To train a reliable reasoner, we collect table knowledge by guiding a teacher LLM to follow the coarse-to-fine reasoning paths and refine it through two quality enhancement strategies to selectively distill the high-quality knowledge to the reasoner. Extensive experiments on two table summarization datasets, including our newly proposed InsTaSumm, validate the general effectiveness of our framework.

SeongKu Kang, Dongha Lee, Wonbin Kweon et al.

Over the past decades, for One-Class Collaborative Filtering (OCCF), many learning objectives have been researched based on a variety of underlying probabilistic models. From our analysis, we observe that models trained with different OCCF objectives capture distinct aspects of user-item relationships, which in turn produces complementary recommendations. This paper proposes a novel OCCF framework, named ConCF, that exploits the complementarity from heterogeneous objectives throughout the training process, generating a more generalizable model. ConCF constructs a multi-branch variant of a given target model by adding auxiliary heads, each of which is trained with heterogeneous objectives. Then, it generates consensus by consolidating the various views from the heads, and guides the heads based on the consensus. The heads are collaboratively evolved based on their complementarity throughout the training, which again results in generating more accurate consensus iteratively. After training, we convert the multi-branch architecture back to the original target model by removing the auxiliary heads, thus there is no extra inference cost for the deployment. Our extensive experiments on real-world datasets demonstrate that ConCF significantly improves the generalization of the model by exploiting the complementarity from heterogeneous objectives.

Dongha Lee, Jiaming Shen, SeongKu Kang et al.

Topic taxonomies, which represent the latent topic (or category) structure of document collections, provide valuable knowledge of contents in many applications such as web search and information filtering. Recently, several unsupervised methods have been developed to automatically construct the topic taxonomy from a text corpus, but it is challenging to generate the desired taxonomy without any prior knowledge. In this paper, we study how to leverage the partial (or incomplete) information about the topic structure as guidance to find out the complete topic taxonomy. We propose a novel framework for topic taxonomy completion, named TaxoCom, which recursively expands the topic taxonomy by discovering novel sub-topic clusters of terms and documents. To effectively identify novel topics within a hierarchical topic structure, TaxoCom devises its embedding and clustering techniques to be closely-linked with each other: (i) locally discriminative embedding optimizes the text embedding space to be discriminative among known (i.e., given) sub-topics, and (ii) novelty adaptive clustering assigns terms into either one of the known sub-topics or novel sub-topics. Our comprehensive experiments on two real-world datasets demonstrate that TaxoCom not only generates the high-quality topic taxonomy in terms of term coherency and topic coverage but also outperforms all other baselines for a downstream task.

Dongha Lee, Dongmin Hyun, Jiawei Han et al.

Identifying outlier documents, whose content is different from the majority of the documents in a corpus, has played an important role to manage a large text collection. However, due to the absence of explicit information about the inlier (or target) distribution, existing unsupervised outlier detectors are likely to make unreliable results depending on the density or diversity of the outliers in the corpus. To address this challenge, we introduce a new task referred to as out-of-category detection, which aims to distinguish the documents according to their semantic relevance to the inlier (or target) categories by using the category names as weak supervision. In practice, this task can be widely applicable in that it can flexibly designate the scope of target categories according to users' interests while requiring only the target-category names as minimum guidance. In this paper, we present an out-of-category detection framework, which effectively measures how confidently each document belongs to one of the target categories based on its category-specific relevance score. Our framework adopts a two-step approach; (i) it first generates the pseudo-category label of all unlabeled documents by exploiting the word-document similarity encoded in a text embedding space, then (ii) it trains a neural classifier by using the pseudo-labels in order to compute the confidence from its target-category prediction. The experiments on real-world datasets demonstrate that our framework achieves the best detection performance among all baseline methods in various scenarios specifying different target categories.

Dongha Lee, Su Kim, Seonghyeon Lee et al.

With the great success of deep learning in various domains, graph neural networks (GNNs) also become a dominant approach to graph classification. By the help of a global readout operation that simply aggregates all node (or node-cluster) representations, existing GNN classifiers obtain a graph-level representation of an input graph and predict its class label using the representation. However, such global aggregation does not consider the structural information of each node, which results in information loss on the global structure. Particularly, it limits the discrimination power by enforcing the same weight parameters of the classifier for all the node representations; in practice, each of them contributes to target classes differently depending on its structural semantic. In this work, we propose structural semantic readout (SSRead) to summarize the node representations at the position-level, which allows to model the position-specific weight parameters for classification as well as to effectively capture the graph semantic relevant to the global structure. Given an input graph, SSRead aims to identify structurally-meaningful positions by using the semantic alignment between its nodes and structural prototypes, which encode the prototypical features of each position. The structural prototypes are optimized to minimize the alignment cost for all training graphs, while the other GNN parameters are trained to predict the class labels. Our experimental results demonstrate that SSRead significantly improves the classification performance and interpretability of GNN classifiers while being compatible with a variety of aggregation functions, GNN architectures, and learning frameworks.

Dongha Lee, Sehun Yu, Hyunjun Ju et al.

Most recent studies on detecting and localizing temporal anomalies have mainly employed deep neural networks to learn the normal patterns of temporal data in an unsupervised manner. Unlike them, the goal of our work is to fully utilize instance-level (or weak) anomaly labels, which only indicate whether any anomalous events occurred or not in each instance of temporal data. In this paper, we present WETAS, a novel framework that effectively identifies anomalous temporal segments (i.e., consecutive time points) in an input instance. WETAS learns discriminative features from the instance-level labels so that it infers the sequential order of normal and anomalous segments within each instance, which can be used as a rough segmentation mask. Based on the dynamic time warping (DTW) alignment between the input instance and its segmentation mask, WETAS obtains the result of temporal segmentation, and simultaneously, it further enhances itself by using the mask as additional supervision. Our experiments show that WETAS considerably outperforms other baselines in terms of the localization of temporal anomalies, and also it provides more informative results than point-level detection methods.

Seonghyeon Lee, Dongha Lee, Hwanjo Yu

Recent studies on neural networks with pre-trained weights (i.e., BERT) have mainly focused on a low-dimensional subspace, where the embedding vectors computed from input words (or their contexts) are located. In this work, we propose a new approach to finding and regularizing the remainder of the space, referred to as out-of-manifold, which cannot be accessed through the words. Specifically, we synthesize the out-of-manifold embeddings based on two embeddings obtained from actually-observed words, to utilize them for fine-tuning the network. A discriminator is trained to detect whether an input embedding is located inside the manifold or not, and simultaneously, a generator is optimized to produce new embeddings that can be easily identified as out-of-manifold by the discriminator. These two modules successfully collaborate in a unified and end-to-end manner for regularizing the out-of-manifold. Our extensive evaluation on various text classification benchmarks demonstrates the effectiveness of our approach, as well as its good compatibility with existing data augmentation techniques which aim to enhance the manifold.

Dongha Lee, SeongKu Kang, Hyunjun Ju et al.

The goal of one-class collaborative filtering (OCCF) is to identify the user-item pairs that are positively-related but have not been interacted yet, where only a small portion of positive user-item interactions (e.g., users' implicit feedback) are observed. For discriminative modeling between positive and negative interactions, most previous work relied on negative sampling to some extent, which refers to considering unobserved user-item pairs as negative, as actual negative ones are unknown. However, the negative sampling scheme has critical limitations because it may choose "positive but unobserved" pairs as negative. This paper proposes a novel OCCF framework, named as BUIR, which does not require negative sampling. To make the representations of positively-related users and items similar to each other while avoiding a collapsed solution, BUIR adopts two distinct encoder networks that learn from each other; the first encoder is trained to predict the output of the second encoder as its target, while the second encoder provides the consistent targets by slowly approximating the first encoder. In addition, BUIR effectively alleviates the data sparsity issue of OCCF, by applying stochastic data augmentation to encoder inputs. Based on the neighborhood information of users and items, BUIR randomly generates the augmented views of each positive interaction each time it encodes, then further trains the model by this self-supervision. Our extensive experiments demonstrate that BUIR consistently and significantly outperforms all baseline methods by a large margin especially for much sparse datasets in which any assumptions about negative interactions are less valid.

Dongha Lee, Seonghyeon Lee, Hwanjo Yu

With the increase of available time series data, predicting their class labels has been one of the most important challenges in a wide range of disciplines. Recent studies on time series classification show that convolutional neural networks (CNN) achieved the state-of-the-art performance as a single classifier. In this work, pointing out that the global pooling layer that is usually adopted by existing CNN classifiers discards the temporal information of high-level features, we present a dynamic temporal pooling (DTP) technique that reduces the temporal size of hidden representations by aggregating the features at the segment-level. For the partition of a whole series into multiple segments, we utilize dynamic time warping (DTW) to align each time point in a temporal order with the prototypical features of the segments, which can be optimized simultaneously with the network parameters of CNN classifiers. The DTP layer combined with a fully-connected layer helps to extract further discriminative features considering their temporal position within an input time series. Extensive experiments on both univariate and multivariate time series datasets show that our proposed pooling significantly improves the classification performance.