Sangkeun Jung

Inbum Heo, Taewook Hwang, Jeesu Jung et al.

Recent advances in Large Language Models (LLMs) and Large Multimodal Models (LMMs) have improved Document Layout Analysis (DLA), yet structural errors such as region merging, splitting, and omission remain persistent. Conventional overlap-based metrics (e.g., IoU, mAP) fail to capture such logical inconsistencies. To overcome this limitation, we propose Layout Error Detection (LED), a benchmark that evaluates structural reasoning in DLA predictions beyond surface-level accuracy. LED defines eight standardized error types (Missing, Hallucination, Size Error, Split, Merge, Overlap, Duplicate, and Misclassification) and provides quantitative rules and injection algorithms for realistic error simulation. Using these definitions, we construct LED-Dataset and design three evaluation tasks: document-level error detection, document-level error-type classification, and element-level error-type classification. Experiments with state-of-the-art multimodal models show that LED enables fine-grained and interpretable assessment of structural understanding, revealing clear weaknesses across modalities and architectures. Overall, LED establishes a unified and explainable benchmark for diagnosing the structural robustness and reasoning capability of document understanding models.

Jong Myoung Kim, Young-Jun Lee, Yong-jin Han et al.

Syntactic elements, such as word order and case markers, are fundamental in natural language processing. Recent studies show that syntactic information boosts language model performance and offers clues for people to understand their learning mechanisms. Unlike languages with a fixed word order such as English, Korean allows for varied word sequences, despite its canonical structure, due to case markers that indicate the functions of sentence components. This study explores whether Korean language models can accurately capture this flexibility. We note that incomplete word orders and omitted case markers frequently appear in ordinary Korean communication. To investigate this further, we introduce the Syntactically Incomplete Korean (SIKO) dataset. Through SIKO, we assessed Korean language models' flexibility with incomplete syntax and confirmed the dataset's training value. Results indicate these models reflect Korean's inherent flexibility, accurately handling incomplete inputs. Moreover, fine-tuning with SIKO enhances the ability to handle common incomplete Korean syntactic forms. The dataset's simple construction process, coupled with significant performance enhancements, solidifies its standing as an effective data augmentation technique.

Hyeonseok Kang, Hyein Seo, Jeesu Jung et al.

While the abundance of rich and vast datasets across numerous fields has facilitated the advancement of natural language processing, sectors in need of specialized data types continue to struggle with the challenge of finding quality data. Our study introduces a novel guidance data augmentation technique utilizing abstracted context and sentence structures to produce varied sentences while maintaining context-entity relationships, addressing data scarcity challenges. By fostering a closer relationship between context, sentence structure, and role of entities, our method enhances data augmentation's effectiveness. Consequently, by showcasing diversification in both entity-related vocabulary and overall sentence structure, and simultaneously improving the training performance of named entity recognition task.

Sungwoo Goo, Hwi-yeol Yun, Sangkeun Jung

For over a decade, explicit memory architectures like the Neural Turing Machine have remained theoretically appealing yet practically intractable for language modeling due to catastrophic gradient instability during Backpropagation Through Time. In this work, we break this stalemate with \textit{Phasor Memory Network} (PMNet), a novel architecture that structurally resolves memory volatility through \textit{Unitary Phasor Dynamics} and \textit{Hierarchical Learnable Anchors}. Rather than relying on brute-force scaling, we present a mechanistic proof-of-concept in a controlled byte-level setting. By constraining recurrent state updates to phase rotations on a complex unit circle, PMNet preserves gradient norms and inherently prevents divergence without the need for specialized initialization. We empirically demonstrate the active actuation of the memory module through a synthetic Copy-Paste task, where PMNet utilizes an expansive \textit{85-slot hierarchical memory tree} ($=\sum^{4}_{h=1}4^{h-1}$) to achieve near 100\% exact retrieval across temporal distances that completely exceed the local sliding window attention's receptive field. Furthermore, despite being a compact 119M parameter model trained on 18.8B tokens, PMNet matches the zero-shot long-context robustness of a Mamba model that is three times larger. Our ablation studies and gradient analyses confirm that the historical failure of explicit memory was a structural alignment problem, which PMNet effectively overcomes, providing a theoretically grounded foundation for scalable sequence modeling.

Hyuk Namgoong, Jeesu Jung, Sangkeun Jung et al.

Recent advancements in large language models have heavily relied on the large reward model from reinforcement learning from human feedback for fine-tuning. However, the use of a single reward model across various domains may not always be optimal, often requiring retraining from scratch when new domain data is introduced. To address these challenges, we explore the utilization of small language models operating in a domain-specific manner based on router mechanisms. Our three approaches are: 1) utilize mixture of experts to form a single reward model by modularizing an internal router and experts, 2) employing external router to select the appropriate reward model from multiple domain-specific models, and 3) the framework reduces parameter size by loading reward models and router adapters onto a single small language model using adapters. Experimental validation underscores the effectiveness of our approach, demonstrating performance comparable to baseline methods while also reducing the total parameter size.

Jeesu Jung, Sangkeun Jung

Curriculum learning for training LLMs requires a difficulty signal that aligns with reasoning while remaining scalable and interpretable. We propose a simple premise: tasks that demand deeper depth of thought for humans should also be harder for models. Accordingly, we define difficulty as depth of thought (DoT) and operationalize it by counting the discrete steps in a teacher model's reasoning trace (e.g., Chain-of-Thought). We then train with a shallow to deep curriculum ordered by this DoT and outline how to derive, validate, and schedule it at scale. Our position yields three testable hypotheses: (i) DoT correlates with conventional difficulty on reasoning benchmarks, (ii) DoT-ordered curricula outperform length- or judge-scored curricula under matched budgets, and (iii) the difficulty is robust across teacher models given light formatting controls. We propose an evaluation framework and discuss threats to validity (teacher style, length confounds) alongside practical mitigations. Taken together, we aim to move toward cognitively grounded, interpretable curricula for reasoning-centric training.

Taewook Hwang, Hyein Seo, Sangkeun Jung

Recent deep learning models such as ChatGPT utilizing the back-propagation algorithm have exhibited remarkable performance. However, the disparity between the biological brain processes and the back-propagation algorithm has been noted. The Forward-Forward algorithm, which trains deep learning models solely through the forward pass, has emerged to address this. Although the Forward-Forward algorithm cannot replace back-propagation due to limitations such as having to use special input and loss functions, it has the potential to be useful in special situations where back-propagation is difficult to use. To work around this limitation and verify usability, we propose an Unsupervised Forward-Forward algorithm. Using an unsupervised learning model enables training with usual loss functions and inputs without restriction. Through this approach, we lead to stable learning and enable versatile utilization across various datasets and tasks. From a usability perspective, given the characteristics of the Forward-Forward algorithm and the advantages of the proposed method, we anticipate its practical application even in scenarios such as federated learning, where deep learning layers need to be trained separately in physically distributed environments.

Goun Pyeon, Inbum Heo, Jeesu Jung et al.

This study systematically evaluated the mathematical reasoning capabilities of Large Language Models (LLMs) using the 2026 Korean College Scholastic Ability Test (CSAT) Mathematics section, ensuring a completely contamination-free evaluation environment. To address data leakage issues in existing benchmarks, we digitized all 46 questions (22 common and 24 elective) within two hours of the exam's public release, eliminating any possibility of inclusion in model training data. We conducted comprehensive evaluations of 24 state-of-the-art LLMs across varying input modalities (Text-only, Image-only, Text+Figure) and prompt languages (Korean, English). The GPT-5 family models achieved perfect scores (100 points) under a limited set of language-modality configurations, while Grok 4, Qwen 3 235B, and Gemini 2.5 pro also scored above 97 points. Notably, gpt-oss-20B achieved 95.7 points despite its relatively small size, demonstrating high cost-effectiveness. Problem-specific analysis revealed Calculus as the weakest domain with significant performance degradation on 4-point high-difficulty problems. Text input consistently outperformed image input, while prompt language effects varied by model scale. In reasoning enhancement experiments with GPT-5 series, increased reasoning intensity improved performance (82.6->100 points) but quadrupled token usage and drastically reduced efficiency, suggesting that models with minimal reasoning may be more practical. This research contributes: (1) implementation of a completely unexposed evaluation environment, (2) a standardized digitization pipeline that converts human-targeted exam materials into LLM-ready evaluation data, and (3) a practical evaluation perspective integrating performance, cost, and time considerations. Detailed results and model comparisons are available at the 2026 Korean CSAT LLM Evaluation Leaderboard; https://isoft.cnu.ac.kr/csat2026/

Inbum Heo, Taewook Hwang, Jeesu Jung et al.

Recent advancements in Document Layout Analysis through Large Language Models and Multimodal Models have significantly improved layout detection. However, despite these improvements, challenges remain in addressing critical structural errors, such as region merging, splitting, and missing content. Conventional evaluation metrics like IoU and mAP, which focus primarily on spatial overlap, are insufficient for detecting these errors. To address this limitation, we propose Layout Error Detection (LED), a novel benchmark designed to evaluate the structural robustness of document layout predictions. LED defines eight standardized error types, and formulates three complementary tasks: error existence detection, error type classification, and element-wise error type classification. Furthermore, we construct LED-Dataset, a synthetic dataset generated by injecting realistic structural errors based on empirical distributions from DLA models. Experimental results across a range of LMMs reveal that LED effectively differentiates structural understanding capabilities, exposing modality biases and performance trade-offs not visible through traditional metrics.

Hyein Seo, Taewook Hwang, Yohan Lee et al.

In English education tutoring, teacher feedback is essential for guiding students. Recently, AI-based tutoring systems have emerged to assist teachers; however, these systems require high-quality and large-scale teacher feedback data, which is both time-consuming and costly to generate manually. In this study, we propose FEAT, a cost-effective framework for generating teacher feedback, and have constructed three complementary datasets: (1) DIRECT-Manual (DM), where both humans and large language models (LLMs) collaboratively generate high-quality teacher feedback, albeit at a higher cost; (2) DIRECT-Generated (DG), an LLM-only generated, cost-effective dataset with lower quality;, and (3) DIRECT-Augmented (DA), primarily based on DG with a small portion of DM added to enhance quality while maintaining cost-efficiency. Experimental results showed that incorporating a small portion of DM (5-10%) into DG leads to superior performance compared to using 100% DM alone.

Jong Myoung Kim, Young-Jun Lee, Ho-Jin Choi et al.

While Large Language Models have gained attention, many service developers still rely on embedding-based models due to practical constraints. In such cases, the quality of fine-tuning data directly impacts performance, and English datasets are often used as seed data for training non-English models. In this study, we propose LANGALIGN, which enhances target language processing by aligning English embedding vectors with those of the target language at the interface between the language model and the task header. Experiments on Korean, Japanese, and Chinese demonstrate that LANGALIGN significantly improves performance across all three languages. Additionally, we show that LANGALIGN can be applied in reverse to convert target language data into a format that an English-based model can process.

Jong Myoung Kim, Young-Jun_Lee, Ho-Jin Choi et al.

Transfer learning leverages the abundance of English data to address the scarcity of resources in modeling non-English languages, such as Korean. In this study, we explore the potential of Phrase Aligned Data (PAD) from standardized Statistical Machine Translation (SMT) to enhance the efficiency of transfer learning. Through extensive experiments, we demonstrate that PAD synergizes effectively with the syntactic characteristics of the Korean language, mitigating the weaknesses of SMT and significantly improving model performance. Moreover, we reveal that PAD complements traditional data construction methods and enhances their effectiveness when combined. This innovative approach not only boosts model performance but also suggests a cost-efficient solution for resource-scarce languages.

Jeesu Jung, Chanjun Park, Sangkeun Jung

Recent efforts in LLM alignment have focused on constructing large-scale preference datasets via human or Artificial Intelligence (AI) annotators. However, such approaches rely on instance-wise supervision, incurring substantial annotation cost and limited interpretability. In this paper, we propose ZEBRA - a model behavior-wise zero-annotation framework that constructs preference data by leveraging model behavior knowledge derived from benchmark performances. ZEBRA binarizes response pairs by evaluating the quality and similarity of their origin models, entirely bypassing instance-level annotation. This allows scalable, controllable, and cost-effective alignment data generation. Empirical results show that ZEBRA achieves alignment performance comparable to instance-supervised methods, despite requiring no manual or model-based labeling.

Hwaran Lee, Seokhwan Jo, HyungJun Kim et al.

The recent advent of neural approaches for developing each dialog component in task-oriented dialog systems has remarkably improved, yet optimizing the overall system performance remains a challenge. Besides, previous research on modeling complicated multi-domain goal-oriented dialogs in end-to-end fashion has been limited. In this paper, we present an effective multi-domain end-to-end trainable neural dialog system SUMBT+LaRL that incorporates two previous strong models and facilitates them to be fully differentiable. Specifically, the SUMBT+ estimates user-acts as well as dialog belief states, and the LaRL models latent system action spaces and generates responses given the estimated contexts. We emphasize that the training framework of three steps significantly and stably increase dialog success rates: separately pretraining the SUMBT+ and LaRL, fine-tuning the entire system, and then reinforcement learning of dialog policy. We also introduce new reward criteria of reinforcement learning for dialog policy training. Then, we discuss experimental results depending on the reward criteria and different dialog evaluation methods. Consequently, our model achieved the new state-of-the-art success rate of 85.4% on corpus-based evaluation, and a comparable success rate of 81.40% on simulator-based evaluation provided by the DSTC8 challenge. To our best knowledge, our work is the first comprehensive study of a modularized E2E multi-domain dialog system that learning from each component to the entire dialog policy for task success.