Akarsh Kumar

Contact

akarshkumar0101(at)gmail(dot)com

Bio
I'm a Ph.D. student at MIT CSAIL working with Phillip Isola.
I'm also a research intern at Sakana AI working with Yujin Tang and David Ha.
I collaborate with Ken Stanley, Jeff Clune, and Joel Lehman.
My research is supported by the NSF GRFP!
Previously, I graduated from UT Austin where I worked with Peter Stone, Risto Miikkulainen, and Alexander Huth.

Research
I want to understand intelligence and all the different processes that emerge it.
More specifically, I'm interested in:

Applying principles from natural evolution and artificial life to create better AI systems.
Open-ended processes which keep creating "interesting" artifacts indefinitely.
Evolving intelligence from scratch, without the internet, like how natural evolution created us.

In practice, I'm especially interested in meta-learning, RL, automatic environment generation, and multi-agent self-play.

Research

FER: Fractured Entangled Representation Hypothesis

Akarsh Kumar, Jeff Clune, Joel Lehman, Kenneth O. Stanley

arXiv 2025

TLDR: Modern deep learning produces good output behavior, but spaghetti-like internal representations. Picbreeder is an example of an open-ended evolutionary algorithm which produces networks with qualitatively more organized internal representations.

[Arxiv] [Code] [Tweet] [MLST Teaser] [MLST Podcast]

ASAL: Automating the Search for Artificial Life with Foundation Models

Akarsh Kumar, Chris Lu, Louis Kirsch, Yujin Tang, Kenneth O. Stanley, Phillip Isola, David Ha

ALife Journal 2025 (Best Oral at ALIFE 2025)

TLDR: ASAL is a new ALife paradigm using VLMs to search for (1) target, (2) open-ended, and (3) diverse simulations. ASAL helps us understand "life as it could be" in arbitrary artificial worlds.

[Web Paper] [Arxiv] [Blog] [Code] [Tweet]

Learning In-Context Decision Making with Synthetic MDPs

Akarsh Kumar, Chris Lu, Louis Kirsch, Phillip Isola

AutoRL @ ICML 2024

TLDR: Generalist in-context RL agents trained on only synthetic MDPs generalize to real MDPs.

[Paper]

GESMR: Effective Mutation Rate Adaptation through Group Elite Selection

Akarsh Kumar, Bo Liu, Risto Miikkulainen, Peter Stone

GECCO 2022

TLDR: Explains why genetic algorithms fail to self-adapt their mutation rate and proposes group selection to fix it.

[Arxiv] [Code]

Physically Plausible Pose Refinement using Fully Differentiable Forces

Akarsh Kumar, Aditya R. Vaidya, Alexander G. Huth

EPIC @ CVPR 2021

TLDR: Accurately refining pose estimations by differentiably modeling the physics of the scene.

[Arxiv] [Presentation]

Talks

Towards a Platonic Intelligence with Unified Factored Representations
Michael Levin's Platonic Space Symposium, November 2025
[Video]

Emergence and Complexity in Artificial Life
Machine Learning Street Talk, October 2025
[Video (Coming Soon)]

FER, Imposter Intelligence, and What's Next
Machine Learning Street Talk, June 2025
[Teaser Video] [Podcast Video]

The Fractured Entangled Representation Hypothesis
Sakana AI, August 2025

Automating the Search for Artificial Life with Foundation Models
ALIFE Oral Presentation, October 2025
[Video (Coming Soon)]

Automating the Search for Artificial Life with Foundation Models
Speaker at Detection and Emergence of Complexity Conference, May 2025
[Video] [DEMECO Website]

Automating the Search for Artificial Life with Foundation Models
MIT Embodied Intelligence Seminar, March 2025
[Video] [Seminar Website]

Automating the Search for Artificial Life with Foundation Models
Kaiming He's Lab, March 2025

Automating the Search for Artificial Life with Foundation Models
Michael Levin's Lab, April 2025
[Video]

Automating the Search for Artificial Life with Foundation Models
Jeff Gore's Lab, March 2025

Learning In-Context Decision Making with Synthetic MDPs
AutoRL @ ICML 2024, May 2024
[Video]