Ayano Hiranaka

I am a CS PhD student at University of Southern California (USC) co-advised by Professor Daniel Seita and Professor Erdem Biyik.

Prior to coming to USC, I completed my Master's degree at Stanford University, where I was a research assistant at Stanford Vision and Learning Lab (SVL). I received my undergraduate degree in mechanical engineering from the University of Illinois at Urbana-Champaign (UIUC).

I have also been fortunate to work as a research intern at Sony AI's Deep Generative Modeling team and Toyota Research Institute's Human Interactive Driving Team.

My experiences are a unique blend of computer science and mechanical engineering, ranging from AI to robotics to mechanical design (my unpublished projects can be found here).

Email  /  CV  /  Google Scholar  /  Github

Publications

I design AI agents, both embodied and non-embodied, that co-evolve with humans through effective communication and collaboration. My research focuses on building human-AI teams that foster mutual improvement by adapting to and learning from one another. Recently, I have been working on developing an AI-assistant that resolve human misconceptions and building a system that streamlines communication and learning between a human user and vision-language-action (VLA) models.

Diagnosing and correcting misconceptions underlying student mistakes to help humans learn generalizable skills for long-horizon decision-making tasks.

Improving robot trajectories by learning reward functions aligned with human preferences, utilizing a joint represntation space of robot trajectory and natural language feedback.

Finetuning text-to-image diffusion models for a variety of tasks in a human-feedback-efficient manner by combining feedback-aligned representation learning and feedback-guided image generation. Work during internship at Sony AI.

Brain-robot interface system for everyday activities using EEG signal decoding, primitive skills, and robot intelligence aided by foundation models.

Combining intuitive skill-based action space and human evaluative feedback, enabling a more safe and sample efficient long-horizon task learning in the real world.

A sample-efficient RLHF framework for low-level robot control policy leveraging a human-interpretable high-level state representation for active query.

Industry Experiences

I have been fortunate to work as a research intern at the following companies:

Toyota Research Institute (Cambridge, 2026): Current intern with the Human Interactive Driving team, mentored by Dr. Deepak Gopinath and Dr. Guy Rosman.

Sony AI (Tokyo, 2024): Developed an RLHF framework for finetuning text-to-image diffusion models, mentored by Dr. Chieh-Hsin Lai in the Deep Generative Modeling team, and Prof. Shao-Hua Sun. Work published at ICLR 2025.

Service

Reviewing: I review for ICLR, ICML, CoRL, ICRA, and workshops at RSS.

Mentoring: I help organize and serve as mentor for the USC CS Undergraduate Mentorship Program that prepare undergraduate students for research and career in CS.