RSI Arena
Can AI agents train a model that wins in
human evaluation
?
Collaborators
Project site
Connecting
Connecting…
Example questions for the final human evaluation
Q
—
is training
—
VS
RSI Arena
Agent trajectory
Budget
—
AI spend
API charges only
—
GPU time
Waiting for GPU usage
Training
Not reported yet
Activity
Tool calls over the session
Original records
Close
Agent trajectory
Execution records