Sitemap
A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.
Pages
publications
Hi-DARTS: Hierarchical Dynamically Adapting Reinforcement Trading System
Published in International Conference on ICT Convergence (ICTC 2025), 2025
Conventional autonomous trading systems struggle to balance computational efficiency and market responsiveness due to their fixed operating frequency. We propose Hi-DARTS, a hierarchical multi-agent reinforcement learning framework in which a meta-agent monitors market volatility and activates frequency-specialized sub-agents, outperforming buy-and-hold baselines in backtesting.
Recommended citation: Hoon Sagong, Heesu Kim, and Hanbeen Hong. (2025). "Hi-DARTS: Hierarchical Dynamically Adapting Reinforcement Trading System." International Conference on ICT Convergence (ICTC 2025).
Download Paper
Cultural Evaluation of Large Language Models
Submitted to ACL Rolling Review (August 2026 cycle)
We extend cultural-bias benchmarks for LLMs — previously limited to Japanese — to a multilingual setting. I led the construction of the Korean subset and evaluated model behavior under zero-shot prompting, chain-of-thought prompting, and supervised fine-tuning. (First author)
Retrieval Without Relevance: Prompt Dependence in Small Machine Translation Models
Published in BabyLM Workshop at EMNLP 2026, 2026
Trained SLMs from scratch under strict rule set to measure how RAG affects machine translation. Compared the machine translation performance across several conditions such as SFT and RAG.
Recommended citation: Heesu Kim, Shota Sato, and Mamoru Komachi. (2026). "Retrieval Without Relevance: Prompt Dependence in Small Machine Translation Models." BabyLM Workshop at EMNLP 2026.
talks
Grounding Word Senses in Vision: Multimodal Word Sense Induction
Published:
Poster presentation on grounding word sense induction in vision: building visual anchors from ImageNet images and using them to disambiguate word senses, comparing text-only and multimodal clustering across SemCor, DWUG EN, and SemEval benchmarks. Joint work with Zhidong Ling, Hajime Kiyama, and Prof. Mamoru Komachi (Komachi Lab, Hitotsubashi University).