Soyeon Kim

Soyoung An, Kyunghoon Bae, Eunbi Choi et al.

We introduce EXAONE 3.0 instruction-tuned language model, the first open model in the family of Large Language Models (LLMs) developed by LG AI Research. Among different model sizes, we publicly release the 7.8B instruction-tuned model to promote open research and innovations. Through extensive evaluations across a wide range of public and in-house benchmarks, EXAONE 3.0 demonstrates highly competitive real-world performance with instruction-following capability against other state-of-the-art open models of similar size. Our comparative analysis shows that EXAONE 3.0 excels particularly in Korean, while achieving compelling performance across general tasks and complex reasoning. With its strong real-world effectiveness and bilingual proficiency, we hope that EXAONE keeps contributing to advancements in Expert AI. Our EXAONE 3.0 instruction-tuned model is available at https://huggingface.co/LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct.

Soyeon Kim, Seongwoo Lim, Kyowoon Lee et al.

Integrated Gradients (IG) is a widely adopted feature attribution method that satisfies desirable axiomatic properties. However, the choice of integration path significantly affects the quality of attributions, and the standard straight-line path introduces all input features simultaneously, often accumulating noisy gradients along the way. To address this limitation, we propose Spectral Integrated Gradients, which constructs integration paths based on singular value decomposition (SVD) of the baseline-to-input difference. By progressively activating singular components from largest to smallest, SIG introduces global structure before fine-grained details, naturally following a coarse-to-fine progression. Through extensive evaluation across diverse image classification datasets, we demonstrate that SIG produces cleaner attribution maps with reduced noise and achieves improved quantitative performance compared to existing path-based attribution methods. Our code is available at https://github.com/leekwoon/sig/.

Kahee Lim, Soyeon Kim, Steven Euijong Whang

As large language models (LLMs) are increasingly deployed in real-world applications, ensuring their fair responses across demographics has become crucial. Despite many efforts, an ongoing challenge is hidden bias: LLMs appear fair under standard evaluations, but can produce biased responses outside those evaluation settings. In this paper, we identify framing -- differences in how semantically equivalent prompts are expressed (e.g., "A is better than B" vs. "B is worse than A") -- as an underexplored contributor to this gap. We first introduce the concept of "framing disparity" to quantify the impact of framing on fairness evaluation. By augmenting fairness evaluation benchmarks with alternative framings, we find that (1) fairness scores vary significantly with framing and (2) existing debiasing methods improve overall (i.e., frame-averaged) fairness, but often fail to reduce framing-induced disparities. To address this, we propose a framing-aware debiasing method that encourages LLMs to be more consistent across framings. Experiments demonstrate that our approach reduces overall bias and improves robustness against framing disparities, enabling LLMs to produce fairer and more consistent responses.

Soyeon Kim, Seongwoo Lim, Kyowoon Lee et al.

Feature attribution is central to diagnosing and trusting deep neural networks, and Integrated Gradients (IG) is widely used due to its axiomatic properties. However, IG can yield unreliable explanations when the integration path between a baseline and the input passes through regions with noisy gradients. While Guided Integrated Gradients reduces this sensitivity by adaptively updating low-gradient-magnitude features, input-space guidance still produces intermediate inputs that deviate from the data manifold. To address this limitation, we propose \emph{Manifold-Aligned Guided Integrated Gradients} (MA-GIG), which constructs attribution paths in the latent space of a pre-trained variational autoencoder. By decoding intermediate latent states, MA-GIG biases the path toward the learned generative manifold and reduces exposure to implausible input-space regions. Through qualitative and quantitative evaluations, we demonstrate that MA-GIG produces faithful explanations by aggregating gradients on path features proximal to the input. Consequently, our method reduces off-manifold noise and outperforms prior path-based attribution methods across multiple datasets and classifiers. Our code is available at https://github.com/leekwoon/ma-gig/.

LG AI Research, Soyoung An, Kyunghoon Bae et al.

This technical report introduces the EXAONE 3.5 instruction-tuned language models, developed and released by LG AI Research. The EXAONE 3.5 language models are offered in three configurations: 32B, 7.8B, and 2.4B. These models feature several standout capabilities: 1) exceptional instruction following capabilities in real-world scenarios, achieving the highest scores across seven benchmarks, 2) outstanding long-context comprehension, attaining the top performance in four benchmarks, and 3) competitive results compared to state-of-the-art open models of similar sizes across nine general benchmarks. The EXAONE 3.5 language models are open to anyone for research purposes and can be downloaded from https://huggingface.co/LGAI-EXAONE. For commercial use, please reach out to the official contact point of LG AI Research: contact_us@lgresearch.ai.

LG AI Research, Kyunghoon Bae, Eunbi Choi et al.

We present EXAONE Deep series, which exhibits superior capabilities in various reasoning tasks, including math and coding benchmarks. We train our models mainly on the reasoning-specialized dataset that incorporates long streams of thought processes. Evaluation results show that our smaller models, EXAONE Deep 2.4B and 7.8B, outperform other models of comparable size, while the largest model, EXAONE Deep 32B, demonstrates competitive performance against leading open-weight models. All EXAONE Deep models are openly available for research purposes and can be downloaded from https://huggingface.co/LGAI-EXAONE

Hyeonbin Hwang, Soyeon Kim, Wei-Jin Park et al.

When it comes to wild conditions, Facial Expression Recognition is often challenged with low-quality data and imbalanced, ambiguous labels. This field has much benefited from CNN based approaches; however, CNN models have structural limitation to see the facial regions in distant. As a remedy, Transformer has been introduced to vision fields with global receptive field, but requires adjusting input spatial size to the pretrained models to enjoy their strong inductive bias at hands. We herein raise a question whether using the deterministic interpolation method is enough to feed low-resolution data to Transformer. In this work, we propose a novel training framework, Neural Resizer, to support Transformer by compensating information and downscaling in a data-driven manner trained with loss function balancing the noisiness and imbalance. Experiments show our Neural Resizer with F-PDLS loss function improves the performance with Transformer variants in general and nearly achieves the state-of-the-art performance.

LG AI Research, Kyunghoon Bae, Eunbi Choi et al.

This technical report introduces EXAONE 4.0, which integrates a Non-reasoning mode and a Reasoning mode to achieve both the excellent usability of EXAONE 3.5 and the advanced reasoning abilities of EXAONE Deep. To pave the way for the agentic AI era, EXAONE 4.0 incorporates essential features such as agentic tool use, and its multilingual capabilities are extended to support Spanish in addition to English and Korean. The EXAONE 4.0 model series consists of two sizes: a mid-size 32B model optimized for high performance, and a small-size 1.2B model designed for on-device applications. The EXAONE 4.0 demonstrates superior performance compared to open-weight models in its class and remains competitive even against frontier-class models. The models are publicly available for research purposes and can be easily downloaded via https://huggingface.co/LGAI-EXAONE.

Soyeon Kim, Cheongwoong Kang, Myeongjin Lee et al.

The development of practical (multimodal) large language model assistants for Korean weather forecasters is hindered by the absence of a multidimensional, expert-level evaluation framework grounded in authoritative sources. To address this, we introduce K-MetBench, a diagnostic benchmark grounded in national qualification exams. It exposes critical gaps across four dimensions: expert visual reasoning of charts, logical validity via expert-verified rationales, Korean-specific geo-cultural comprehension, and fine-grained domain analysis. Our evaluation of 55 models reveals a profound modality gap in interpreting specialized diagrams and a reasoning gap where models hallucinate logic despite correct predictions. Crucially, Korean models outperform significantly larger global models in local contexts, demonstrating that parameter scaling alone cannot resolve cultural dependencies. K-MetBench serves as a roadmap for developing reliable, culturally aware expert AI agents. The dataset is available at https://huggingface.co/datasets/soyeonbot/K-MetBench .

Eunbi Choi, Kibong Choi, Seokhee Hong et al.

This technical report presents K-EXAONE, a large-scale multilingual language model developed by LG AI Research. K-EXAONE is built on a Mixture-of-Experts architecture with 236B total parameters, activating 23B parameters during inference. It supports a 256K-token context window and covers six languages: Korean, English, Spanish, German, Japanese, and Vietnamese. We evaluate K-EXAONE on a comprehensive benchmark suite spanning reasoning, agentic, general, Korean, and multilingual abilities. Across these evaluations, K-EXAONE demonstrates performance comparable to open-weight models of similar size. K-EXAONE, designed to advance AI for a better life, is positioned as a powerful proprietary AI foundation model for a wide range of industrial and research applications.

Eunbi Choi, Kibong Choi, Sehyun Chun et al.

This technical report introduces EXAONE 4.5, the first open-weight vision language model released by LG AI Research. EXAONE 4.5 is architected by integrating a dedicated visual encoder into the existing EXAONE 4.0 framework, enabling native multimodal pretraining over both visual and textual modalities. The model is trained on large-scale data with careful curation, particularly emphasizing document-centric corpora that align with LG's strategic application domains. This targeted data design enables substantial performance gains in document understanding and related tasks, while also delivering broad improvements across general language capabilities. EXAONE 4.5 extends context length up to 256K tokens, facilitating long-context reasoning and enterprise-scale use cases. Comparative evaluations demonstrate that EXAONE 4.5 achieves competitive performance in general benchmarks while outperforming state-of-the-art models of similar scale in document understanding and Korean contextual reasoning. As part of LG's ongoing effort toward practical industrial deployment, EXAONE 4.5 is designed to be continuously extended with additional domains and application scenarios to advance AI for a better life.

Soyeon Kim, Jindong Wang, Xing Xie et al.

Facts evolve over time, making it essential for Large Language Models (LLMs) to handle time-sensitive factual knowledge accurately and reliably. While factual Time-Sensitive Question-Answering (TSQA) tasks have been widely studied, existing benchmarks often rely on manual curation or a small, fixed set of predefined templates, which restricts scalable and comprehensive TSQA evaluation. To address these challenges, we propose TDBench, a new benchmark that systematically constructs TSQA pairs by harnessing temporal databases and database techniques such as temporal SQL and functional dependencies. We also introduce a fine-grained evaluation metric called time accuracy, which assesses the validity of time references in model explanations alongside traditional answer accuracy to enable a more reliable TSQA evaluation. Extensive experiments on contemporary LLMs show how \ours{} enables scalable and comprehensive TSQA evaluation while reducing the reliance on human labor, complementing existing Wikipedia/Wikidata-based TSQA evaluation approaches by enabling LLM evaluation on application-specific data and seamless multi-hop question generation. Code and data are publicly available at: https://github.com/ssoy0701/tdbench.git.

Dongjun Jang, Sangah Lee, Sungjoo Byun et al.

This paper presents the DaG LLM (David and Goliath Large Language Model), a language model specialized for Korean and fine-tuned through Instruction Tuning across 41 tasks within 13 distinct categories.

Dongkeun Yoon, Seungone Kim, Sohee Yang et al. · cmu, deepmind

Despite their strengths, large language models (LLMs) often fail to communicate their confidence accurately, making it difficult to assess when they might be wrong and limiting their reliability. In this work, we demonstrate that reasoning models that engage in extended chain-of-thought (CoT) reasoning exhibit superior performance not only in problem-solving but also in accurately expressing their confidence. Specifically, we benchmark six reasoning models across six datasets and find that they achieve strictly better confidence calibration than their non-reasoning counterparts in 33 out of the 36 settings. Our detailed analysis reveals that these gains in calibration stem from the slow thinking behaviors of reasoning models (e.g., exploring alternative approaches and backtracking) which enable them to adjust their confidence dynamically throughout their CoT, making it progressively more accurate. In particular, we find that reasoning models become increasingly better calibrated as their CoT unfolds, a trend not observed in non-reasoning models. Moreover, removing slow thinking behaviors from the CoT leads to a significant drop in calibration. Lastly, we show that non-reasoning models also demonstrate enhanced calibration when simply guided to slow think via in-context learning, fully isolating slow thinking as the source of the calibration gains.

Boseon Yoo, Jiwoo Lee, Janghoon Ju et al.

We introduce a novel loss function, Covariance Loss, which is conceptually equivalent to conditional neural processes and has a form of regularization so that is applicable to many kinds of neural networks. With the proposed loss, mappings from input variables to target variables are highly affected by dependencies of target variables as well as mean activation and mean dependencies of input and target variables. This nature enables the resulting neural networks to become more robust to noisy observations and recapture missing dependencies from prior information. In order to show the validity of the proposed loss, we conduct extensive sets of experiments on real-world datasets with state-of-the-art models and discuss the benefits and drawbacks of the proposed Covariance Loss.

Jio Oh, Soyeon Kim, Junseok Seo et al.

Large language models (LLMs) have achieved unprecedented performances in various applications, yet evaluating them is still challenging. Existing benchmarks are either manually constructed or are automatic, but lack the ability to evaluate the thought process of LLMs with arbitrary complexity. We contend that utilizing existing relational databases based on the entity-relationship (ER) model is a promising approach for constructing benchmarks as they contain structured knowledge that can be used to question LLMs. Unlike knowledge graphs, which are also used to evaluate LLMs, relational databases have integrity constraints that can be used to better construct complex in-depth questions and verify answers: (1) functional dependencies can be used to pinpoint critical keywords that an LLM must know to properly answer a given question containing certain attribute values; and (2) foreign key constraints can be used to join relations and construct multi-hop questions, which can be arbitrarily long and used to debug intermediate answers. We thus propose ERBench, which uses these integrity constraints to convert any database into an LLM benchmark. ERBench supports continuous evaluation as databases change, multimodal questions, and various prompt engineering techniques. In our experiments, we construct LLM benchmarks using databases of multiple domains and make an extensive comparison of contemporary LLMs. We show how ERBench can properly evaluate any LLM by not only checking for answer correctness, but also effectively verifying the rationales by looking for the right keywords.

Soyeon Kim, Junho Choi, Yeji Choi et al.

Machine learning (ML) is becoming increasingly popular in meteorological decision-making. Although the literature on explainable artificial intelligence (XAI) is growing steadily, user-centered XAI studies have not extend to this domain yet. This study defines three requirements for explanations of black-box models in meteorology through user studies: statistical model performance for different rainfall scenarios to identify model bias, model reasoning, and the confidence of model outputs. Appropriate XAI methods are mapped to each requirement, and the generated explanations are tested quantitatively and qualitatively. An XAI interface system is designed based on user feedback. The results indicate that the explanations increase decision utility and user trust. Users prefer intuitive explanations over those based on XAI algorithms even for potentially easy-to-recognize examples. These findings can provide evidence for future research on user-centered XAI algorithms, as well as a basis to improve the usability of AI systems in practice.

Seokhee Hong, Sunkyoung Kim, Guijin Son et al. · allen-ai, anthropic

The development of Large Language Models (LLMs) requires robust benchmarks that encompass not only academic domains but also industrial fields to effectively evaluate their applicability in real-world scenarios. In this paper, we introduce two Korean expert-level benchmarks. KMMLU-Redux, reconstructed from the existing KMMLU, consists of questions from the Korean National Technical Qualification exams, with critical errors removed to enhance reliability. KMMLU-Pro is based on Korean National Professional Licensure exams to reflect professional knowledge in Korea. Our experiments demonstrate that these benchmarks comprehensively represent industrial knowledge in Korea. We release our dataset publicly available.

Subeen Lee, Jiyeon Han, Soyeon Kim et al.

Deep generative models are proficient in generating realistic data but struggle with producing rare samples in low density regions due to their scarcity of training datasets and the mode collapse problem. While recent methods aim to improve the fidelity of generated samples, they often reduce diversity and coverage by ignoring rare and novel samples. This study proposes a novel approach for generating diverse rare samples from high-resolution image datasets with pretrained GANs. Our method employs gradient-based optimization of latent vectors within a multi-objective framework and utilizes normalizing flows for density estimation on the feature space. This enables the generation of diverse rare images, with controllable parameters for rarity, diversity, and similarity to a reference image. We demonstrate the effectiveness of our approach both qualitatively and quantitatively across various datasets and GANs without retraining or fine-tuning the pretrained GANs.

Soyeon Kim, Namhee Kim, Yeonwoo Jeong

Effective evaluation of web data record extraction methods is crucial, yet hampered by static, domain-specific benchmarks and opaque scoring practices. This makes fair comparison between traditional algorithmic techniques, which rely on structural heuristics, and Large Language Model (LLM)-based approaches, offering zero-shot extraction across diverse layouts, particularly challenging. To overcome these limitations, we introduce a concrete evaluation framework. Our framework systematically generates evaluation datasets from arbitrary MHTML snapshots, annotates XPath-based supervision labels, and employs structure-aware metrics for consistent scoring, specifically preventing text hallucination and allowing only for the assessment of positional hallucination. It also incorporates preprocessing strategies to optimize input for LLMs while preserving DOM semantics: HTML slimming, Hierarchical JSON, and Flat JSON. Additionally, we created a publicly available synthetic dataset by transforming DOM structures and modifying content. We benchmark deterministic heuristic algorithms and off-the-shelf LLMs across these multiple input formats. Our benchmarking shows that Flat JSON input enables LLMs to achieve superior extraction accuracy (F1 score of 0.9567) and minimal hallucination compared to other input formats like Slimmed HTML and Hierarchical JSON. We establish a standardized foundation for rigorous benchmarking, paving the way for the next principled advancements in web data record extraction.

Soyeon Kim, Junho Choi, Subeen Lee et al.

To improve the trustworthiness of an AI model, finding consistent, understandable representations of its inference process is essential. This understanding is particularly important in high-stakes operations such as weather forecasting, where the identification of underlying meteorological mechanisms is as critical as the accuracy of the predictions. Despite the growing literature that addresses this issue through explainable AI, the applicability of their solutions is often limited due to their AI-centric development. To fill this gap, we follow a user-centric process to develop an example-based concept analysis framework, which identifies cases that follow a similar inference process as the target instance in a target model and presents them in a user-comprehensible format. Our framework provides the users with visually and conceptually analogous examples, including the probability of concept assignment to resolve ambiguities in weather mechanisms. To bridge the gap between vector representations identified from models and human-understandable explanations, we compile a human-annotated concept dataset and implement a user interface to assist domain experts involved in the the framework development.

Xiyuan Yang, Shengyuan Hu, Soyeon Kim et al.

Federated clustering aims to group similar clients into clusters and produce one model for each cluster. Such a personalization approach typically improves model performance compared with training a single model to serve all clients, but can be more vulnerable to privacy leakage. Directly applying client-level differentially private (DP) mechanisms to federated clustering could degrade the utilities significantly. We identify that such deficiencies are mainly due to the difficulties of averaging privacy noise within each cluster (following standard privacy mechanisms), as the number of clients assigned to the same clusters is uncontrolled. To this end, we propose a simple and effective technique, named RR-Cluster, that can be viewed as a light-weight add-on to many federated clustering algorithms. RR-Cluster achieves reduced privacy noise via randomly rebalancing cluster assignments, guaranteeing a minimum number of clients assigned to each cluster. We analyze the tradeoffs between decreased privacy noise variance and potentially increased bias from incorrect assignments and provide convergence bounds for RR-Clsuter. Empirically, we demonstrate the RR-Cluster plugged into strong federated clustering algorithms results in significantly improved privacy/utility tradeoffs across both synthetic and real-world datasets.

Soyeon Kim, Jihyeon Seong, Hyunkyung Han et al.

Capsule Neural Networks utilize capsules, which bind neurons into a single vector and learn position equivariant features, which makes them more robust than original Convolutional Neural Networks. CapsNets employ an affine transformation matrix and dynamic routing with coupling coefficients to learn robustly. In this paper, we investigate the effectiveness of CapsNets in analyzing highly sensitive and noisy time series sensor data. To demonstrate CapsNets robustness, we compare their performance with original CNNs on electrocardiogram data, a medical time series sensor data with complex patterns and noise. Our study provides empirical evidence that CapsNets function as noise stabilizers, as investigated by manual and adversarial attack experiments using the fast gradient sign method and three manual attacks, including offset shifting, gradual drift, and temporal lagging. In summary, CapsNets outperform CNNs in both manual and adversarial attacked data. Our findings suggest that CapsNets can be effectively applied to various sensor systems to improve their resilience to noise attacks. These results have significant implications for designing and implementing robust machine learning models in real world applications. Additionally, this study contributes to the effectiveness of CapsNet models in handling noisy data and highlights their potential for addressing the challenges of noise data in time series analysis.