Ilia Sucholutsky

Papers

Total Citations

H-Index

About

Ilia Sucholutsky is an emerging researcher working at the intersection of machine learning, human-robot interaction, and representational learning. His work focuses on developing more robust and generalizable methods for how intelligent systems learn from human input — particularly through demonstrations and language guidance. His most notable recent contribution, "Preference-Conditioned Language-Guided Abstraction" (2024), addresses a fundamental challenge in robot learning: when robots learn from human demonstrations, they often pick up on irrelevant visual features, leading to brittle, non-generalizable behavior. Sucholutsky's approach leverages natural language to construct meaningful state abstractions — visual representations that capture only task-relevant information — and further conditions these on user preferences, enabling more personalized and reliable learning. This work reflects a broader commitment to making machine learning systems that are not only technically capable but also responsive to human intent. Though early in citation accumulation with 5 citations, the work tackles a highly relevant problem in an active research space, positioning Sucholutsky as a promising voice in the fields of imitation learning, abstraction theory, and human-aligned AI systems.