https://jungjunkim9150.github.io/Jung-jun KimA minimal, responsive and feature-rich Jekyll theme for technical writing. 2026-07-05T17:54:43+08:00 your_full_name https://jungjunkim9150.github.io/ Jekyll © 2026 your_full_name /assets/img/favicons/favicon.ico /assets/img/favicons/favicon-96x96.png ChartCoder2026-07-04T00:00:00+08:00 2026-07-04T22:13:39+08:00 https://jungjunkim9150.github.io/posts/chartcoder/ jungjunkim ChartCoder: Advancing Multimodal Large Language Model for Chart-to-Code Generation Authors: Xuanle Zhao, Xianzhen Luo, Qi Shi, Chi Chen, Shuo Wang, Wanxiang Che, Zhiyuan Liu, Maosong Sun (Tsinghua University, Harbin Institute of Technology) — arXiv 2501.06598, ACL 2025. 차트→코드(chart-to-code) 논문임. 차트 이미지를 matplotlib 코드로 복원하는 첫 전용(dedicated) MLLM으로, 앞서 본 MSRL의 직계 선행 연구임(1저자 Xuanle Zhao 공통). M... Matryoshka Representation Learning2026-07-02T00:00:00+08:00 2026-07-02T00:00:00+08:00 https://jungjunkim9150.github.io/posts/matryoshka/ jungjunkim Matryoshka Representation Learning (MRL) Authors: Aditya Kusupati, Gantavya Bhatt, Aniket Rege, Matthew Wallingford, Aditya Sinha, Vivek Ramanujan, William Howard-Snyder, Kaifeng Chen, Sham Kakade, Prateek Jain, Ali Farhadi (University of Washington, Google Research, Harvard) — arXiv 2205.13147, NeurIPS 2022. 표현학습(representation learning) 논문임. 특정 task용 모델이 아니라, 이미 있는 임베딩 파이프라인 위에 얹는 범용 기법임... Unified LTR for Multi-Channel Retrieval2026-06-27T00:00:00+08:00 2026-06-27T02:53:48+08:00 https://jungjunkim9150.github.io/posts/unified-ltr-multichannel/ jungjunkim Unified Learning-to-Rank for Multi-Channel Retrieval in Large-Scale E-Commerce Search Authors: Aditya Gaydhani, Guangyue Xu, Dhanush Kamath, Ankit Singh, Alex Li (Target Corporation) — arXiv 2602.23530, 2026 (Preprint). “여러 검색 채널의 결과를 어떻게 하나의 순위로 합치느냐”는 문제를 LTR로 다시 정의한 논문임. 검색 파이프라인 retrieval → ranking → re-ranking 중 마지막 단계(채널 병합/재랭킹)에 위치함. 클릭·장바구니·구매 같은 engagement 신호를 라벨이자 피처로 동시에 쓰는 전형적인... MSRL2026-06-24T00:00:00+08:00 2026-06-24T00:00:00+08:00 https://jungjunkim9150.github.io/posts/msrl/ jungjunkim Breaking the SFT Plateau: Multimodal Structured Reinforcement Learning for Chart-to-Code Generation Authors: Lei Chen, Xuanle Zhao, Zhixiong Zeng, Jing Huang, Liming Zheng, Yufeng Zhong, Lin Ma (Meituan) — arXiv 2508.13587, ICLR 2026. 모델명 MSRL. 차트→코드 생성(chart-to-code) 논문이며, 직전에 본 MOCR과 같은 “이미지를 실행 가능한 코드로 복원”하는 계보에 있음. 저자진(Xuanle Zhao·Zhixiong Zeng·Yufeng Zhong·Lin Ma 등)이 MOCR 리뷰에서 다룬 OCRV... Beyond Human Annotation2026-06-21T00:00:00+08:00 2026-06-21T00:00:00+08:00 https://jungjunkim9150.github.io/posts/beyond-human-annotation/ jungjunkim Beyond Human Annotation: Recent Advances in Data Generation Methods for Document Intelligence Authors: Dehao Ying, Fengchang Yu, Haihua Chen, Changjiang Jiang, Yurong Li, Wei Lu (Wuhan University, University of North Texas) — arXiv 2601.12318, 2026년 1월. 서베이(survey) 논문임. 모델이 아니라 그 모델들이 먹는 학습 데이터를 어떻게 만드느냐를 다룸. 문서 AI(DI) 전반의 데이터 생성 기법을 기법이 아니라 “데이터·라벨 가용성”이라는 자원(resource) 축으로 처음 재편한 분류 체계임. ...