Ifdita Hasan Orney

I am a CS Master's student at Stanford University. Previously, I did my undergraduate in CS at Stanford University. My research is currently advised by Dorsa Sadigh and Emma Brunskill. I work in artificial intelligence with a focus on reinforcement learning and human-centered AI. Outside of AI research, I love doing dance and art.

Email  /  Github  /  Google Scholar

I am interested in sequential decision making using RL. More specifically, I am interested in exploration and sample-efficient RL fine-tuning. I am also interested in building adaptive, feedback-driven models with an eye on applications such as in education and personalized learning.

(*) denotes corresponding author

• Poly-EPO: Training Exploratory Reasoning Models.
  Ifdita Hasan Orney*, Jubayer Ibn Hamid*, Shreya Ramanujam, Shirley Wu, Hengyuan Hu, Noah Goodman, Dorsa Sadigh, Chelsea Finn.
  In Submission
  (Paper)

• Polychromic Objectives for Reinforcement Learning.
  Jubayer Ibn Hamid*, Ifdita Hasan Orney*, Ellen Xu, Chelsea Finn, Dorsa Sadigh.
  International Conference on Learning Representations (ICLR), 2026.
  (Paper)

• Can LLM-Simulated Practice and Feedback Upskill Human Counselors? A Randomized Study with 90+ Novice Counselors.
  Ryan Louie*, Ifdita Hasan Orney, Juan Pablo Pacheco, Raj Sanjay Shah, Emma Brunskill, Diyi Yang.
  Proceedings of the CHI Conference on Human Factors in Computing Systems, 2026.
  (Website) (Paper)
  (Honorable Mention Award)

Projects

A Critical Study of the Entropy Bonus for Exploration

Critical study of how entropy impacts policy exploration and critic stability during RLFT.
(Code) (Paper)
(Nominated for Best Paper - Stanford's CS224R)

ELI5B: Explain Like I'm 5B

Improving Agentic LLM–SLM Communication Through the Llama 4 Herd.
(Code)
(Honorable Mention - Meta 8VC Hackathon)

Applications

CARECoach

Build core skills for peer support (Website)