Jeonghoon Kim

Publications

*equal contribution

Lead author

1. 1
   NeurIPS 2026

   A Simple Unigram Cross-Entropy Lens on the Lexical Imprint of Pre-training Data (opens in a new tab)

   Jeonghoon Kim*, Woojin Chung*, Woomin Song, Cheonbok Park, Nako Sung, Jinwoo Shin

   • Paper: A Simple Unigram Cross-Entropy Lens on the Lexical Imprint of Pre-training Data (opens in a new tab)
   • arXiv (earlier version): A Simple Unigram Cross-Entropy Lens on the Lexical Imprint of Pre-training Data (opens in a new tab)
2. 2
   ICML 2025

   Peri-LN: Revisiting Normalization Layer in the Transformer Architecture (opens in a new tab)

   Jeonghoon Kim, Byeongchan Lee, Cheonbok Park, Yeontaek Oh, Beomjun Kim, Taehwan Yoo, Seongjin Shin, Dongyoon Han, Jinwoo Shin, Kang Min Yoo

   • Paper: Peri-LN: Revisiting Normalization Layer in the Transformer Architecture (opens in a new tab)
   • arXiv: Peri-LN: Revisiting Normalization Layer in the Transformer Architecture (opens in a new tab)
3. 3
   NeurIPS 2025

   Exploiting Vocabulary Frequency Imbalance in Language Model Pre-training (opens in a new tab)

   Woojin Chung*, Jeonghoon Kim*

   • Paper: Exploiting Vocabulary Frequency Imbalance in Language Model Pre-training (opens in a new tab)
   • arXiv: Exploiting Vocabulary Frequency Imbalance in Language Model Pre-training (opens in a new tab)
   • Code: Exploiting Vocabulary Frequency Imbalance in Language Model Pre-training (opens in a new tab)
4. 4
   NAACL 2025

   LRQ: Optimizing Post-Training Quantization for Large Language Models by Learning Low-Rank Weight-Scaling Matrices (opens in a new tab)

   Jung Hyun Lee*, Jeonghoon Kim*, June Yong Yang, Se Jung Kwon, Eunho Yang, Kang Min Yoo, Dongsoo Lee

   • Paper: LRQ: Optimizing Post-Training Quantization for Large Language Models by Learning Low-Rank Weight-Scaling Matrices (opens in a new tab)
   • arXiv: LRQ: Optimizing Post-Training Quantization for Large Language Models by Learning Low-Rank Weight-Scaling Matrices (opens in a new tab)
5. 5
   ICLR 2024 Mentorship program

   Rethinking Channel Dimensions to Isolate Outliers for Low-bit Weight Quantization of Large Language Models (opens in a new tab)

   Jung Hwan Heo*, Jeonghoon Kim*, Beomseok Kwon, Byeongwook Kim, Se Jung Kwon, Dongsoo Lee

   • Paper: Rethinking Channel Dimensions to Isolate Outliers for Low-bit Weight Quantization of Large Language Models (opens in a new tab)
   • arXiv: Rethinking Channel Dimensions to Isolate Outliers for Low-bit Weight Quantization of Large Language Models (opens in a new tab)
   • Code: Rethinking Channel Dimensions to Isolate Outliers for Low-bit Weight Quantization of Large Language Models (opens in a new tab)
6. 6
   NeurIPS 2023

   Memory-Efficient Fine-Tuning of Compressed Large Language Models via sub-4-bit Integer Quantization (opens in a new tab)

   Jeonghoon Kim*, Jung Hyun Lee*, Sungdong Kim, Joonsuk Park, Kang Min Yoo, Se Jung Kwon, Dongsoo Lee

   • Paper: Memory-Efficient Fine-Tuning of Compressed Large Language Models via sub-4-bit Integer Quantization (opens in a new tab)
   • arXiv: Memory-Efficient Fine-Tuning of Compressed Large Language Models via sub-4-bit Integer Quantization (opens in a new tab)
7. 7
   ICML 2023

   FlexRound: Learnable Rounding based on Element-wise Division for Post-Training Quantization (opens in a new tab)

   Jung Hyun Lee*, Jeonghoon Kim*, Se Jung Kwon, Dongsoo Lee

   • Paper: FlexRound: Learnable Rounding based on Element-wise Division for Post-Training Quantization (opens in a new tab)
   • arXiv: FlexRound: Learnable Rounding based on Element-wise Division for Post-Training Quantization (opens in a new tab)
8. 8
   Arxiv (preprint, 2025)

   Cross-lingual Collapse: How Language-Centric Foundation Models Shape Reasoning in Large Language Models (opens in a new tab)

   Cheonbok Park*, Jeonghoon Kim*, Joosung Lee, Sanghwan Bae, Jaegul Choo, Kang Min Yoo

   • arXiv: Cross-lingual Collapse: How Language-Centric Foundation Models Shape Reasoning in Large Language Models (opens in a new tab)

Co-author

9. 9
   Technical report (Arxiv, 2026) Lead of the Omni-pathfinding mission.

   HyperCLOVA X 8B Omni (opens in a new tab)

   HyperCLOVA X Team

   • arXiv: HyperCLOVA X 8B Omni (opens in a new tab)
   • Model: HyperCLOVA X 8B Omni (opens in a new tab)
10. 10
    Technical report (Arxiv, 2026)

    HyperCLOVA X 32B Think (opens in a new tab)

    HyperCLOVA X Team

    • arXiv: HyperCLOVA X 32B Think (opens in a new tab)
11. 11
    EMNLP 2026

    OmniACBench: A Benchmark for Evaluating Context-Grounded Acoustic Control in Omni-Modal Models (opens in a new tab)

    Seunghee Kim, Bumkyu Park, Kyudan Jung, Joosung Lee, Soyoon Kim, Jeonghoon Kim, Taeuk Kim, Hwiyeol Jo

    • arXiv: OmniACBench: A Benchmark for Evaluating Context-Grounded Acoustic Control in Omni-Modal Models (opens in a new tab)
12. 12
    Findings of EMNLP 2026

    What Models Know, How Well They Know It: Knowledge-Weighted Fine-Tuning for Learning When to Say "I Don't Know" (opens in a new tab)

    Joosung Lee, Hwiyeol Jo, Donghyeon Ko, Kyubyung Chae, Cheonbok Park, Jeonghoon Kim

    • arXiv: What Models Know, How Well They Know It: Knowledge-Weighted Fine-Tuning for Learning When to Say "I Don't Know" (opens in a new tab)
13. 13
    Interspeech 2026

    ParaPairAudioBench: Paralinguistic Pairwise Audio Benchmark for LALM-as-a-Judge (opens in a new tab)

    Jisu Jeon, Seungyeon Jwa, Joosung Lee, Jinhyeon Kim, Woojin Chung, Hwiyeol Jo, Jeonghoon Kim, Jonghyun Choi, Soyoon Kim

    • Paper: ParaPairAudioBench: Paralinguistic Pairwise Audio Benchmark for LALM-as-a-Judge (opens in a new tab)
    • arXiv: ParaPairAudioBench: Paralinguistic Pairwise Audio Benchmark for LALM-as-a-Judge (opens in a new tab)
14. 14
    ACL 2026 (Industry Track)

    Sommelier: Scalable Open Multi-turn Audio Pre-processing for Full-duplex Speech Language Models (opens in a new tab)

    Kyudan Jung, Jihwan Kim, Soyoon Kim, Jeonghoon Kim, Jaegul Choo, Cheonbok Park

    • Paper: Sommelier: Scalable Open Multi-turn Audio Pre-processing for Full-duplex Speech Language Models (opens in a new tab)
    • arXiv: Sommelier: Scalable Open Multi-turn Audio Pre-processing for Full-duplex Speech Language Models (opens in a new tab)
15. 15
    Findings of ACL 2026 Co-corresponding author.

    ReflexiCoder: Teaching Large Language Models to Self-Reflect on Generated Code and Self-Correct It via Reinforcement Learning (opens in a new tab)

    Juyong Jiang, Jiasi Shen, Sunghun Kim, Kang Min Yoo, Jeonghoon Kim, Sungju Kim

    • Paper: ReflexiCoder: Teaching Large Language Models to Self-Reflect on Generated Code and Self-Correct It via Reinforcement Learning (opens in a new tab)
    • arXiv: ReflexiCoder: Teaching Large Language Models to Self-Reflect on Generated Code and Self-Correct It via Reinforcement Learning (opens in a new tab)
    • Code: ReflexiCoder: Teaching Large Language Models to Self-Reflect on Generated Code and Self-Correct It via Reinforcement Learning (opens in a new tab)
16. 16
    Findings of ACL 2026

    Enhancing Hallucination Detection via Future Context (opens in a new tab)

    Joosung Lee, Cheonbok Park, Hwiyeol Jo, Jeonghoon Kim, Joonsuk Park, Kang Min Yoo

    • Paper: Enhancing Hallucination Detection via Future Context (opens in a new tab)
    • arXiv: Enhancing Hallucination Detection via Future Context (opens in a new tab)
17. 17
    Arxiv (preprint, 2025)

    ReGUIDE: Data Efficient GUI Grounding via Spatial Reasoning and Search (opens in a new tab)

    Hyunseok Lee, Jeonghoon Kim, Beomjun Kim, Jihoon Tack, Chansong Jo, Jaehong Lee, Cheonbok Park, Sookyo In, Jinwoo Shin, Kang Min Yoo

    • arXiv: ReGUIDE: Data Efficient GUI Grounding via Spatial Reasoning and Search (opens in a new tab)
18. 18
    Technical report (Arxiv, 2025) Uses the Peri-LN architecture.

    HyperCLOVA X THINK Technical Report (opens in a new tab)

    HyperCLOVA X Team

    • arXiv: HyperCLOVA X THINK Technical Report (opens in a new tab)
19. 19
    ICLR 2024

    LUT-GEMM: Quantized Matrix Multiplication based on LUTs for Efficient Inference in Large-Scale Generative Language Models (opens in a new tab)

    Gunho Park, Baeseong Park, Minsub Kim, Sungjae Lee, Jeonghoon Kim, Beomseok Kwon, Se Jung Kwon, Byeongwook Kim, Youngjoo Lee, Dongsoo Lee

    • Paper: LUT-GEMM: Quantized Matrix Multiplication based on LUTs for Efficient Inference in Large-Scale Generative Language Models (opens in a new tab)
    • arXiv: LUT-GEMM: Quantized Matrix Multiplication based on LUTs for Efficient Inference in Large-Scale Generative Language Models (opens in a new tab)
    • Code: LUT-GEMM: Quantized Matrix Multiplication based on LUTs for Efficient Inference in Large-Scale Generative Language Models (opens in a new tab)
20. 20
    Technical report (Arxiv, 2024)

    HyperCLOVA X Technical Report (opens in a new tab)

    HyperCLOVA X Team

    • arXiv: HyperCLOVA X Technical Report (opens in a new tab)
21. 21
    ICLR 2023

    Winning Both the Accuracy of Floating Point Activation and the Simplicity of Integer Arithmetic (opens in a new tab)

    Yulhwa Kim, Jaeyong Jang, Jehun Lee, Jihoon Park, Jeonghoon Kim, Byeongwook Kim, Baeseong Park, Se Jung Kwon, Dongsoo Lee, Jae-Joon Kim

    • Paper: Winning Both the Accuracy of Floating Point Activation and the Simplicity of Integer Arithmetic (opens in a new tab)
22. 22
    Findings of EMNLP 2022

    AlphaTuning: Quantization-Aware Parameter-Efficient Adaptation of Large-Scale Pre-Trained Language Models (opens in a new tab)

    Se Jung Kwon, Jeonghoon Kim, Jeongin Bae, Kang Min Yoo, Jin-Hwa Kim, Baeseong Park, Byeongwook Kim, Jung-Woo Ha, Nako Sung, Dongsoo Lee

    • Paper: AlphaTuning: Quantization-Aware Parameter-Efficient Adaptation of Large-Scale Pre-Trained Language Models (opens in a new tab)
    • arXiv: AlphaTuning: Quantization-Aware Parameter-Efficient Adaptation of Large-Scale Pre-Trained Language Models (opens in a new tab)