research
Papers and ongoing work.
-
ENCORE: Few-Shot Agentic Discovery of Manipulation Strategies
A coding agent reads three demonstrations, writes an executable policy program, and repairs it from its own execution traces. Oral, NeurIPS 2026 Robot Learning Workshop.
-
Interpreting Diffusion Language Models Over Denoising Time
Sparse autoencoders and cross-layer transcoders for diffusion LLMs, with timestep-aware tracing and causal interventions. CoLM 2026 Workshop on Non-Autoregressive LMs.
-
Doodle Agent: Exploring Freeform Visual Generation with Multimodal LLMs
A drawing agent for open-ended visual creation. Co-first author. ICCV 2025 AI4VA Workshop.
-
CNOT-Optimal Circuit Synthesis
Written at the Research Science Institute (RSI); selected as a top-5 paper out of 100.