publications
-
Which State Should Prediction Read? A Mechanistic Analysis of State–Prediction Separation
ICLR 2027 Under reviewUnder review at ICLR 2027
@misc{kim2026sps, title = {Which State Should Prediction Read? A Mechanistic Analysis of State--Prediction Separation}, author = {Kim, Ryan}, year = {2026}, howpublished = {Under review at ICLR 2027}, } -
From Text to Treatment: Automated Identification of Airway Events and Referral Patterns in Lung Cancer Imaging Using Locally Deployed CPU-Only LLMs
ACRO 2026 Annual Meeting Accepted abstractBuilt a CPU-only LLM pipeline (LLaMA3.1-8B, Qwen3-4B with DSPy/MIPROv2) to classify lung cancer CT reports by specialty referral category, analyzing 370 clinically annotated reports at MSK.
@misc{kim2026acro, title = {From Text to Treatment: Automated Identification of Airway Events and Referral Patterns in Lung Cancer Imaging Using Locally Deployed CPU-Only LLMs}, author = {Amoako-Boadu, Kwadwo and Kim, Alexander and Kim, Ryan and Miller, David and Naaman, Rami and Book, Abraham J. and Lee, Robert and Shaverdian, Narek and Gomez, Daniel and Chaunzwa, Tafadzwa L.}, year = {2026}, note = {Abstract accepted at ACRO 2026 Annual Meeting. Memorial Sloan Kettering Cancer Center \& Weill Cornell Medicine.}, } -
Contrastive fine-tuning against stain and scanner bias improves robustness and downstream performance across pathology foundation models
NeurIPS 2026 Workshop AcceptedNeurIPS 2026 Workshop on AI at Scale for Clinical Impact: Cancer Pathology Foundation Models (ASCI)
@inproceedings{kim2026contrastive, title = {Contrastive fine-tuning against stain and scanner bias improves robustness and downstream performance across pathology foundation models}, author = {Kim, Ryan and Muhammad, Hassan and Dulal, Anish and Scotti, Paul Steven and Abraham, Tanishq Mathew}, booktitle = {NeurIPS 2026 Workshop on AI at Scale for Clinical Impact: Cancer Pathology Foundation Models (ASCI)}, year = {2026}, note = {Accepted.}, } -
NanoPath: Fast, Fair, Open Experimentation for Pathology Foundation Models
NeurIPS 2026 Workshop AcceptedNeurIPS 2026 Workshop on AI at Scale for Clinical Impact: Cancer Pathology Foundation Models (ASCI)
@inproceedings{scotti2026nanopath, title = {{NanoPath}: Fast, Fair, Open Experimentation for Pathology Foundation Models}, author = {Scotti, Paul Steven and Muhammad, Hassan and Kim, Ryan and Ashjaee, Nima and Grandhi, Ratna Sagari and Rahbar, Saman and Biju, Raymond and Dulal, Anish and Feng, Chao and Lane, Connor and Warner, Benjamin and Abraham, Tanishq Mathew}, booktitle = {NeurIPS 2026 Workshop on AI at Scale for Clinical Impact: Cancer Pathology Foundation Models (ASCI)}, year = {2026}, note = {Accepted.}, }