ASTRA: Automated Synthesis of agentic Trajectories and Reinforcement Arenas

Open paper
Benchmark Model Rank Results
agents-on-benchASTRA-32B-thinking-v1#23Accuracy: 63.7
mathematical-reasoning-on-aime24ASTRA-32B-Thinking-v1#25Acc: 81.4

Rank counts only results with a code link.