“Executable Counterfactuals: Improving LLMs’ Causal Reasoning Through Code” accepted to NeurIPS 2025 Workshop on Foundations of Reasoning in Language Models (co-first author).
“The Best Instruction-Tuning Data are Those That Fit” accepted to NeurIPS 2025 (Spotlight).
“Improving Influence-based Instruction Tuning Data Selection for Balanced Learning of Diverse Capabilities” accepted to Findings of EMNLP 2025 and ICLR 2025 Workshop on DATA-FM.
Presenting GRAPE (Spotlight) and Executable Counterfactuals at NeurIPS 2025 (December 2025).
Presented BIDS work at the 2nd DATA-FM Workshop, ICLR 2025 (April 2025).