Scholar
Guijin Son
Google Scholar ID: Zf_eLDsAAAAJ
Undergraduate, Yonsei University
Citations & Impact
Citations
275
H-index
9
i10-index
9
Publications
20
Co-authors
5
list available
Publications
Resume (English only)
Academic Achievements
- 2025: Papers 'KMMLU-Pro' and 'Multi-LMentry' accepted to EMNLP 2025
- 2025: 'Linguistic Generalizability' (Oral) and 'FinKRX' (Industry Track) accepted to ACL 2025
- 2025: 'Robustness of Reward Models' accepted to ICML 2025
- 2025: 'BiGGen Bench' and 'KMMLU' accepted to NAACL 2025
- 2025: 'BiGGen Bench' awarded Best Paper at NAACL 2025
- 2024: 'Multitask Inference' accepted to ACL 2024
- 2024: 'HAE-RAE Bench' accepted to LREC-COLING 2024
- Multiple preprints under review on multilingual reasoning, scientific verification benchmark (SPOT), and multilingual meta-evaluation (MM-Eval)
Background
- Co-Founder at OneLine AI
- Lead of HAE-RAE, an open-source research group focused on Korean NLP
- Research interests: AI for Science, evaluation and reasoning with language models, multimodal reasoning, and agentic systems
- Current goal: building stronger reasoning models and developing metrics to demonstrate real progress
- Past projects include Korean knowledge and professional benchmarks, reward model evaluation, and financial applications of LLMs
- Teaches at Fast Campus and SSAFY; contributes to curriculum development with Codeit and Code States; mentors at Upstage
