👥 Looking for teammates?

MeltingPot Challenge 2023-ca9cce
#1

Competing is more fun with a team!

Introduce yourself here, and find others who are looking to team up! :sparkles:

:writing_hand: Format:

  • A short introduction about you and your background.
  • What brings you to this challenge?
  • Some ideas you wish to explore as a part of this challenge?

Cheers,
Team MeltingPot

1 Like
#2
#3
#4
#5

A short introduction about you and your background:

Hi everyone! I’m Nicolo’, a PhD student in Computer Science at Sapienza University of Rome. My research is focused on Reinforcement Learning in the context of Natural Language Processing.

What brings you to this challenge?
I’m particularly interested on communication for coordination in mixed human-robot teams and what insights we can gain from fully artificial systems.

Some ideas you wish to explore as part of this challenge?

  1. Emergent Communication: exploring how communication can emerge autonomously between agents.
  2. Human-AI Collaboration: Learning about coordination MAS systems to improve mixed human robot teams coordination.
#6

A short introduction about you and your background.
Hi Guys. I am Viswanathan. I am a Graduate student from India focusing on Multi Agent RL for High Power Laser Systems.

What brings you to this challenge?
Looking to work with passionate people on such exciting projects.

Some ideas you wish to explore as part of this challenge?
Self-Play and Competitive Training:

In some cases, agents can be trained initially in competitive settings to learn diverse strategies.
They can then be fine-tuned in cooperative settings.

#7
  • A short introduction about you and your background:
    I am in China and very interesting in Computer vision, ML, AI as well. but now I am not very good at computer vision. Now I am working on Medical system development, and familiar with Python, Numpy, Pandas as well. I wish we have chance to learn each other and make progress with the challenge.
  • What brings you to this challenge:
    there are a lot of trade-off or tricks could be designed and shared cross agents to archive the optimized goal, I may think of we have to figure out some rule-based measurable features to evaluate dynamically performance and then reward or punish with.
  • Some ideas you wish to explore as a part of this challenge?
  1. clarify the goal and measurable points in advance
  2. figure out features stright-forward to state the behavior clearly
  3. startup with supervised learning with partial instance of training, then gradually extend to reinforcement learning
#8

hey everyone!

I am Sacha, 2nd year mater student from Moscow (MIPT). Trying my best in doing research in multi-agent systems, especially in cooperative intelligence. Done some prior work in natural language processing and educational projects in RL.

What brings you to this challenge?
I’m interested in communication in cooperative intelligence and its game theoretic explanations. In particular, hope to get into coding proof algorithms to help understand social interactions.

Some ideas you wish to explore as part of this challenge?

  1. Efficient pipeline: write high performant code in Jax/Flax to parallelise it/vectorise it
  2. Emergent communication: endue emergent communication between agents to provoke stronger communication protocols
  3. Training Curriculum: how to train such systems to be generalisable to many substrates/scenarios, maybe metalearning

I do not have computational resources but I can code and provide a lot of ideas. Pretty disciplined and responsible as well xD

#9

Hey Nicole! I’m Alif, a sophomore CS major from Cornell University. I’m pretty new to the field of reinforcement learning, and this project does seem a bit advanced for beginners, if I’m being honest, but I’m willing to take the deep plunge and learn what I need to. I would be excited to partner with you!

#10

About Me:
I’m a recent Computer Science graduate from Cleveland State University with experience in Large Language Models (LLMs) and a passion for AI. When I’m not coding, you’ll find me immersed in role-playing games, anime, or books.

Why I’m Here:
I’m here to further my skills in AI, especially within multi-agent reinforcement learning. The opportunity to understand LLMs in decision-making and the potential of AI sublanguages intrigues me.

Ideas I Want to Explore:
I’m keen to investigate how LLMs can enhance decision-making in cooperative scenarios and how agents balance individual and group goals when faced with new partners.

If you share similar interests, let’s collaborate and push the boundaries of AI together!

#11

A short introduction about you and your background:
Hey all!
My name is Max, I’m a recent applied math grad from Columbia University, looking to get some experience in multi-agent RL ahead of a PhD in computational cognitive science. Mainly my experience is in mathematics, NLP, and the broader data science stack, in academia and industry.
What brings you to this challenge?
I am very interested in human-AI interaction and building human-compatible AI systems. I strongly believe multi-agent coordination is key to building safe systems in the future, and I aim to bring a computational approach to modeling and improving collaborative, communicative interactions.
Some ideas you wish to explore as part of this challenge?

  1. simulating social inference/reference games with many agents with varying levels of recall, truthfulness, and deceptive intent
  2. performance of human-ai vs. ai-ai teams in collaborative games with varying reward structures
  3. conditions that lead to manipulative/antisocial/disassociative behavior in a group setting.

Feel free to reach out if you’re interested in working together - looking forward to coding!

#12

Hey Alex! do you have a team for the hackathon yet? I’d be interested in teaming. Let me know! - Max

#13

unfortunately, still struggle
you can shoot me an email at work dot alexch2 at gmail dot com

#14

Hi everyone! I’m Anna a computational neuroscience and cs student from Technion. I have experience in RL and NLP and my research is now focused on multi agent rl especially social learning

What brings you to this challenge?

I am interesed interesed in multi agent rl from a game theoretic and cognitive science perspective .Specifically,I am interested in emergent collective behaviour via multi agent rl and if and how communication protocols can be leveraged to encourage cooperative behaviour.

Some ideas I find interesting

  1. social hierarchies and how they can be used as intrinsic rewards to drive social welfare
    2)Also using inverse rl to model theory of mind for indicidual agents/groups seems like a cool direction to go in

Please dont hesitate to reach out if this seems interesting

#15

Hi Anna, I’m Dheeraj an final year IT undergrad.

I’m a beginner, your background is very impressive and a path I want to follow. I think I can learn so much from you.
I’m interested in LLM engineering and NLP eng.
If you have a place in your team, you can consider me.

Thank you.