Jinyeop Song

sjinyeop@gmail.com / Google Scholar / GitHub / LinkedIn / X

I am a final-year PhD student in the MIT Physics Department, advised by Prof. Jeff Gore. I am a former MATS scholar and MIT-IBM Watson AI Lab research intern.

I work on the Science of AI and AI for Science. My recent work studies neural scaling laws, in-context learning, and LLM agents, while also building foundation-model and agentic systems for biology.

Selected publications

• Estimating the Empowerment of Language Model Agents. Jinyeop Song, Jeff Gore, Max Kleiman-Weiner. ICML 2026 / Paper
• The Blessing of Dimensionality in LLM Fine-tuning: A Variance-Curvature Perspective. Qiyao Liang, Jinyeop Song, Yizhou Liu, Jeff Gore, Ila R. Fiete, Risto Miikkulainen, Xin Qiu. Submitted 2026 / Paper
• Efficient and Transferable Agentic Knowledge Graph RAG via Reinforcement Learning. Jinyeop Song, Song Wang, Julian Shun, Yada Zhu. Under review 2025 / PDF
• Emergence and Effectiveness of Task Vectors in In-Context Learning: An Encoder Decoder Perspective. Seungwook Han, Jinyeop Song, Jeff Gore, Pulkit Agrawal. ICML 2025 Spotlight / Paper
• Reconciling Kaplan and Chinchilla Scaling Laws. Tim Pearce, Jinyeop Song. TMLR 2024 / Paper
• MethylGPT: a foundation model for the DNA methylome. Kejun Ying, Jinyeop Song, Haotian Cui, et al. bioRxiv 2024 / Paper

Awards and honors

• Machine Learning and Theory Scholarship (MATS), 2025
• Mokam Research Scholarship, 2023
• KPS Best Oral Presentation Award, 2023
• Asan Foundation Biomedical Science Scholarship, 2021-2025
• Excellent Graduate of KAIST, 2020
• KAIST Presidential Fellowship (KPF), 2016
• Korea Presidential Science Scholarship, 2014-2020

Projects

• Protein LLM: adapting open-source LLMs to protein modality via post-training.
• Autoresearch De Novo: LLM agents for inference-time protein design optimization.

See publications for a fuller list.