Ayush Jain

I am a Computer Science PhD student in at University of Southern California, co-advised by Prof. Joseph J. Lim and Prof. Erdem Bıyık. I was fortunate to intern at Meta Reality Labs, Microsoft Research Montreal and Naver AI, Seoul. Before joining USC, I spent two years in Seoul, working at Samsung Research Korea. Earlier, I graduated from IIT Delhi, where I worked under the guidance of Prof. Sumeet Agarwal and Prof. Rajakrishnan Rajkumar.

Research Goal: To build adaptive and capable agents for both physical and virtual worlds. I work on reinforcement learning algorithms and architectures that learn under complex action spaces that are large, unseen, varying, or difficult to optimize.

Email  |  Twitter  |  CV  |  Google Scholar  |  LinkedIn

Research
SAVO Teaser Mitigating Suboptimality of Deterministic Policy Gradients in Complex Q-functions
Ayush Jain, Norio Kosaka, Xinhu Li, Kyung-Min Kim, Erdem Bıyık, Joseph J. Lim
preprint

We identify that TD3 gets stuck in local optima in tasks with complex Q-functions and propose a new actor architecture to find better optima.

arXiv

QMP Teaser QMP: Q-switch Mixture of Policies for Multi-Task Behavior Sharing
Grace Zhang*, Ayush Jain*, Injune Hwang, Shao-Hua Sun, Joseph J. Lim
preprint

Sharing behaviors between tasks to improve exploration for multitask reinforcement learning.

arXiv | Project Page

Know Your Action Set: Learning Action Relations for Reinforcement Learning
Ayush Jain*, Norio Kosaka*, Kyung-Min Kim, Joseph J. Lim
International Conference on Learning Representations (ICLR), 2022

For optimal decision-making under a varying action space, we learn the relations between the available actions using a graph-attention network based policy architecture.

Paper | Project Page | Code | Talk

Generalization to New Actions in Reinforcement Learning
Ayush Jain*, Andrew Szot*, Joseph J. Lim
International Conference on Machine Learning (ICML), 2020

Our proposed RL framework enables agents to solve sequential decision-making tasks even when the available actions (tools or skills) have not been seen before.

Paper | Project Page | arXiv | Code | Talk | Environment

UID Uniform Information Density Effects on Syntactic Choice in Hindi
Ayush Jain*, Vishal Singh*, Sidharth Ranjan*, Rajakrishnan Rajkumar, Sumeet Agarwal
Workshop on Linguistic Complexity and Natural Language Processing, COLING 2018

This work investigates the extent to which word order choices in Hindi language are influenced by the drive to minimize the information variance in a sentence.

Teaching

Teaching Assistant (USC): Deep Learning and its Applications (CSCI566, CSCI599)

  • Fall 2024: Prof. Yan Liu
  • Spring 2024: Prof. Yue Zhao
  • Spring 2023: Prof. Jesse Thomason
  • Fall 2020: Prof. Joseph J Lim
  • Spring 2019: Prof. Joseph J Lim
  • Fall 2019: Prof. Joseph J Lim

Reviewing
  • ICLR: 2023, 2024, 2025
  • NeurIPS: 2023, 2024
  • CoRL: 2021, 2022, 2023, 2024

Credits to the Coolest template!