DigiRL-agent / digirl

Official repo for paper DigiRL: Training In-The-Wild Device-Control Agents with Autonomous Reinforcement Learning.
Apache License 2.0
200 stars 16 forks source link

Compare DigiRL to Agent-Q #15

Closed James4Ever0 closed 3 weeks ago

James4Ever0 commented 1 month ago

Agent Q is the latest reinforcement learning web agent built by MultiOn. Wonder how DigiRL performs compared to it, in the same context.

Paper: https://arxiv.org/html/2408.07199v1 Blog: https://www.multion.ai/blog/introducing-agent-q-research-breakthrough-for-the-next-generation-of-ai-agents-with-planning-and-self-healing-capabilities

BiEchi commented 4 weeks ago

Thanks for your interest in our work. We're currently developing some better algorithms upon DigiRL and might include the performance of Agent Q in the same context in the new work.

BiEchi commented 3 weeks ago

Closing due to inactivity.