Vanya Cohen

Papers

Total Citations

H-Index

About

Vanya Cohen is a researcher at the intersection of robotics, natural language processing, and artificial intelligence, with a primary focus on enabling robots to reason and recover from failures using large language models (LLMs). Their most cited work, "CAPE: Corrective Actions from Precondition Errors using Large Language Models" (2022, 2024), introduces a novel framework that leverages LLMs to diagnose and correct action failures in robotic planning—moving beyond simple retries to resolve underlying precondition errors. This work, accumulating over 20 citations, addresses a critical gap in LLM-based robotics: the inability to autonomously recover from execution failures. Cohen’s earlier research, "Grounding Language Attributes to Objects using Bayesian Eigenobjects" (2019), developed a system for disambiguating object instances from natural language descriptions and depth images, contributing to the field of grounded language understanding. By combining Bayesian inference with object representations, this work laid groundwork for more nuanced human-robot interaction. Cohen’s contributions are particularly notable for their practical impact on autonomous systems, offering scalable solutions for real-world robotic tasks. Their research continues to shape how robots interpret and act upon human instructions, bridging the gap between high-level language and low-level control.