Joo-Kyung Kim

Taehee Jung, Joo-Kyung Kim, Sungjin Lee et al. · amazon-science

For extreme multi-label classification (XMC), existing classification-based models poorly perform for tail labels and often ignore the semantic relations among labels, like treating "Wikipedia" and "Wiki" as independent and separate labels. In this paper, we cast XMC as a generation task (XLGen), where we benefit from pre-trained text-to-text models. However, generating labels from the extremely large label space is challenging without any constraints or guidance. We, therefore, propose to guide label generation using label cluster information to hierarchically generate lower-level labels. We also find that frequency-based label ordering and using decoding ensemble methods are critical factors for the improvements in XLGen. XLGen with cluster guidance significantly outperforms the classification and generation baselines on tail labels, and also generally improves the overall performance in four popular XMC benchmarks. In human evaluation, we also find XLGen generates unseen but plausible labels. Our code is now available at https://github.com/alexa/xlgen-eacl-2023.

Jinyoung Park, Ameen Patel, Omar Zia Khan et al.

Chain-of-Thought (CoT) prompting along with sub-question generation and answering has enhanced multi-step reasoning capabilities of Large Language Models (LLMs). However, prompting the LLMs to directly generate sub-questions is suboptimal since they sometimes generate redundant or irrelevant questions. To deal with them, we propose a GE-Reasoning method, which directs LLMs to generate proper sub-questions and corresponding answers. Concretely, given an input question, we first prompt the LLM to generate knowledge triplets, forming a graph representation of the question. Unlike conventional knowledge triplets, our approach allows variables as head or tail entities, effectively representing a question as knowledge triplets. Second, for each triplet, the LLM generates a corresponding sub-question and answer along with using knowledge retrieval. If the prediction confidence exceeds a threshold, the sub-question and prediction are incorporated into the prompt for subsequent processing. This approach encourages that sub-questions are grounded in the extracted knowledge triplets, reducing redundancy and irrelevance. Our experiments demonstrate that our approach outperforms previous CoT prompting methods and their variants on multi-hop question answering benchmark datasets.

Zae Myung Kim, Anand Ramachandran, Farideh Tavazoee et al.

Generating long, coherent text remains a challenge for large language models (LLMs), as they lack hierarchical planning and structured organization in discourse generation. We introduce Structural Alignment, a novel method that aligns LLMs with human-like discourse structures to enhance long-form text generation. By integrating linguistically grounded discourse frameworks into reinforcement learning, our approach guides models to produce coherent and well-organized outputs. We employ a dense reward scheme within a Proximal Policy Optimization framework, assigning fine-grained, token-level rewards based on the discourse distinctiveness relative to human writing. Two complementary reward models are evaluated: the first improves readability by scoring surface-level textual features to provide explicit structuring, while the second reinforces deeper coherence and rhetorical sophistication by analyzing global discourse patterns through hierarchical discourse motifs, outperforming both standard and RLHF-enhanced models in tasks such as essay generation and long-document summarization. All training data and code will be publicly shared at https://github.com/minnesotanlp/struct_align.

Soyeong Jeong, Taehee Jung, Sung Ju Hwang et al.

Recent Long-Context Language Models (LCLMs) can process hundreds of thousands of tokens in a single prompt, enabling new opportunities for knowledge-intensive multi-hop reasoning by integrating large sets of retrieved documents or, in some cases, directly all necessary information. However, simply feeding more documents into the context window fails to capture how evidence should be connected. We address this gap with thought templates, which recast reasoning as reusable thought caches, derived from prior problem solving traces, structuring how evidence is combined and guiding multi-hop inference with factual documents. To keep these templates effective, we propose an update strategy that iteratively refines templates derived from training data through natural-language feedback. Across diverse benchmarks and LCLM families, our approach delivers consistent gains over strong baselines in both retrieval-based and retrieval-free settings. Furthermore, we show that optimized templates can be distilled into smaller open-source models, demonstrating its broad applicability and transparent reasoning reuse. We refer to our framework as Thought Template Augmented LCLMs (ToTAL).

Chanwoo Park, Seungju Han, Xingzhi Guo et al. · stanford

Leveraging multiple large language models (LLMs) to build collaborative multi-agentic workflows has demonstrated significant potential. However, most previous studies focus on prompting the out-of-the-box LLMs, relying on their innate capability for collaboration, which may not improve LLMs' performance as shown recently. In this paper, we introduce a new post-training paradigm MAPoRL (Multi-Agent Post-co-training for collaborative LLMs with Reinforcement Learning), to explicitly elicit the collaborative behaviors and further unleash the power of multi-agentic LLM frameworks. In MAPoRL, multiple LLMs first generate their own responses independently and engage in a multi-turn discussion to collaboratively improve the final answer. In the end, a MAPoRL verifier evaluates both the answer and the discussion, by assigning a score that verifies the correctness of the answer, while adding incentives to encourage corrective and persuasive discussions. The score serves as the co-training reward, and is then maximized through multi-agent RL. Unlike existing LLM post-training paradigms, MAPoRL advocates the co-training of multiple LLMs together using RL for better generalization. Accompanied by analytical insights, our experiments demonstrate that training individual LLMs alone is insufficient to induce effective collaboration. In contrast, multi-agent co-training can boost the collaboration performance across benchmarks, with generalization to unseen domains.

Shirley Anugrah Hayati, Taehee Jung, Tristan Bodding-Long et al.

Fine-tuning large language models (LLMs) with a collection of large and diverse instructions has improved the model's generalization to different tasks, even for unseen tasks. However, most existing instruction datasets include only single instructions, and they struggle to follow complex instructions composed of multiple subtasks. In this work, we propose a novel concept of compositional instructions called chain-of-instructions (CoI), where the output of one instruction becomes an input for the next like a chain. Unlike the conventional practice of solving single instruction tasks, our proposed method encourages a model to solve each subtask step by step until the final answer is reached. CoI-tuning (i.e., fine-tuning with CoI instructions) improves the model's ability to handle instructions composed of multiple subtasks as well as unseen composite tasks such as multilingual summarization. Overall, our study find that simple CoI tuning of existing instruction data can provide consistent generalization to solve more complex, unseen, and longer chains of instructions.

Jihyung Kil, Farideh Tavazoee, Dongyeop Kang et al.

Visual Question Answering (VQA) often involves diverse reasoning scenarios across Vision and Language (V&L). Most prior VQA studies, however, have merely focused on assessing the model's overall accuracy without evaluating it on different reasoning cases. Furthermore, some recent works observe that conventional Chain-of-Thought (CoT) prompting fails to generate effective reasoning for VQA, especially for complex scenarios requiring multi-hop reasoning. In this paper, we propose II-MMR, a novel idea to identify and improve multi-modal multi-hop reasoning in VQA. In specific, II-MMR takes a VQA question with an image and finds a reasoning path to reach its answer using two novel language promptings: (i) answer prediction-guided CoT prompt, or (ii) knowledge triplet-guided prompt. II-MMR then analyzes this path to identify different reasoning cases in current VQA benchmarks by estimating how many hops and what types (i.e., visual or beyond-visual) of reasoning are required to answer the question. On popular benchmarks including GQA and A-OKVQA, II-MMR observes that most of their VQA questions are easy to answer, simply demanding "single-hop" reasoning, whereas only a few questions require "multi-hop" reasoning. Moreover, while the recent V&L model struggles with such complex multi-hop reasoning questions even using the traditional CoT method, II-MMR shows its effectiveness across all reasoning cases in both zero-shot and fine-tuning settings.

Jinyoung Park, Minseok Joo, Joo-Kyung Kim et al.

Knowledge graph-grounded dialog generation requires retrieving a dialog-relevant subgraph from the given knowledge base graph and integrating it with the dialog history. Previous works typically represent the graph using an external encoder, such as graph neural networks, and retrieve relevant triplets based on the similarity between single-vector representations of triplets and the dialog history. However, these external encoders fail to leverage the rich knowledge of pretrained language models, and the retrieval process is also suboptimal due to the information bottleneck caused by the single-vector abstraction of the dialog history. In this work, we propose Dialog generation with Generative Subgraph Retrieval (DialogGSR), which retrieves relevant knowledge subgraphs by directly generating their token sequences on top of language models. For effective generative subgraph retrieval, we introduce two key methods: (i) structure-aware knowledge graph linearization with self-supervised graph-specific tokens and (ii) graph-constrained decoding utilizing graph structural proximity-based entity informativeness scores for valid and relevant generative retrieval. DialogGSR achieves state-of-the-art performance in knowledge graph-grounded dialog generation, as demonstrated on OpenDialKG and KOMODIS datasets.

Jinyoung Park, Sanghyeok Lee, Omar Zia Khan et al.

Graph Retrieval-Augmented Generation (GraphRAG) has been successfully applied in various knowledge-intensive question answering tasks by organizing external knowledge into structured graphs of entities and relations. It enables large language models (LLMs) to perform complex reasoning beyond text-chunk retrieval. Recent works have employed reinforcement learning (RL) to train agentic GraphRAG frameworks that perform iterative interactions between LLMs and knowledge graphs. However, existing RL-based frameworks such as Graph-R1 suffer from two key limitations: (1) they primarily depend on semantic similarity for retrieval, often overlooking the underlying graph structure, and (2) they rely on sparse, outcome-level rewards, failing to capture the quality of intermediate retrieval steps and their dependencies. To address these limitations, we propose ProGraph-R1, a progress-aware agentic framework for graph-based retrieval and multi-step reasoning. ProGraph-R1 introduces a structure-aware hypergraph retrieval mechanism that jointly considers semantic relevance and graph connectivity, encouraging coherent traversal along multi-hop reasoning paths. We also design a progress-based step-wise policy optimization, which provides dense learning signals by modulating advantages according to intermediate reasoning progress within a graph, rather than relying solely on final outcomes. Experiments on multi-hop question answering benchmarks demonstrate that ProGraph-R1 consistently improves reasoning accuracy and generation quality over existing GraphRAG methods.

Qiusi Zhan, Angeline Budiman-Chan, Abdelrahman Zayed et al.

Large language model (LLM) based search agents iteratively generate queries, retrieve external information, and reason to answer open-domain questions. While researchers have primarily focused on improving their utility, their safety behaviors remain underexplored. In this paper, we first evaluate search agents using red-teaming datasets and find that they are more likely to produce harmful outputs than base LLMs. For example, when asked "How can I track someone's location without their consent?", a base model refuses, whereas a search agent designed to retrieve and cite sources may lower its refusal threshold, fetch documents (e.g., court cases), and, once appended, synthesize them into an informative yet unsafe summary. We further show that utility-oriented fine-tuning intensifies this risk, motivating joint alignment of safety and utility. We present SafeSearch, a multi-objective reinforcement learning approach that couples a final-output safety/utility reward with a novel query-level shaping term that penalizes unsafe queries and rewards safe ones. Experiments show that SafeSearch reduces agent harmfulness by over 70% across three red-teaming datasets while producing safe, helpful responses, and matches the QA performance of a utility-only finetuned agent; further analyses confirm the effectiveness of the query-level reward in jointly improving safety and utility.

Joo-Kyung Kim, Guoyin Wang, Sungjin Lee et al.

A large-scale conversational agent can suffer from understanding user utterances with various ambiguities such as ASR ambiguity, intent ambiguity, and hypothesis ambiguity. When ambiguities are detected, the agent should engage in a clarifying dialog to resolve the ambiguities before committing to actions. However, asking clarifying questions for all the ambiguity occurrences could lead to asking too many questions, essentially hampering the user experience. To trigger clarifying questions only when necessary for the user satisfaction, we propose a neural self-attentive model that leverages the hypotheses with ambiguities and contextual signals. We conduct extensive experiments on five common ambiguity types using real data from a large-scale commercial conversational agent and demonstrate significant improvement over a set of baseline approaches.

Joo-Kyung Kim, Young-Bum Kim

In large-scale domain classification, an utterance can be handled by multiple domains with overlapped capabilities. However, only a limited number of ground-truth domains are provided for each training utterance in practice while knowing as many as correct target labels is helpful for improving the model performance. In this paper, given one ground-truth domain for each training utterance, we regard domains consistently predicted with the highest confidences as additional pseudo labels for the training. In order to reduce prediction errors due to incorrect pseudo labels, we leverage utterances with negative system responses to decrease the confidences of the incorrectly predicted domains. Evaluating on user utterances from an intelligent conversational system, we show that the proposed approach significantly improves the performance of domain classification with hypothesis reranking.

Joo-Kyung Kim, Young-Bum Kim

In large-scale domain classification for natural language understanding, leveraging each user's domain enablement information, which refers to the preferred or authenticated domains by the user, with attention mechanism has been shown to improve the overall domain classification performance. In this paper, we propose a supervised enablement attention mechanism, which utilizes sigmoid activation for the attention weighting so that the attention can be computed with more expressive power without the weight sum constraint of softmax attention. The attention weights are explicitly encouraged to be similar to the corresponding elements of the ground-truth's one-hot vector by supervised attention, and the attention information of the other enabled domains is leveraged through self-distillation. By evaluating on the actual utterances from a large-scale IPDA, we show that our approach significantly improves domain classification performance.

Joo-Kyung Kim, Young-Bum Kim

In domain classification for spoken dialog systems, correct detection of out-of-domain (OOD) utterances is crucial because it reduces confusion and unnecessary interaction costs between users and the systems. Previous work usually utilizes OOD detectors that are trained separately from in-domain (IND) classifiers, and confidence thresholding for OOD detection given target evaluation scores. In this paper, we introduce a neural joint learning model for domain classification and OOD detection, where dynamic class weighting is used during the model training to satisfice a given OOD false acceptance rate (FAR) while maximizing the domain classification accuracy. Evaluating on two domain classification tasks for the utterances from a large spoken dialogue system, we show that our approach significantly improves the domain classification performance with satisficing given target FARs.

Young-Bum Kim, Dongchan Kim, Joo-Kyung Kim et al.

Intelligent personal digital assistants (IPDAs), a popular real-life application with spoken language understanding capabilities, can cover potentially thousands of overlapping domains for natural language understanding, and the task of finding the best domain to handle an utterance becomes a challenging problem on a large scale. In this paper, we propose a set of efficient and scalable neural shortlisting-reranking models for large-scale domain classification in IPDAs. The shortlisting stage focuses on efficiently trimming all domains down to a list of k-best candidate domains, and the reranking stage performs a list-wise reranking of the initial k-best domains with additional contextual information. We show the effectiveness of our approach with extensive experiments on 1,500 IPDA domains.