Agentic Task Completion on Terminal-Bench Easy 4 tasks 2
100Pass@1TF-GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| TF-GRPOBase Model=GPT-5.4 (high reasoning), Rollouts (k)=22026.04 | 100 | |
| AHEBase Model=GPT-5.4 (high reasoning), Rollouts (k)=22026.04 | 100 | |
| ACEBase Model=GPT-5.4 (high reasoning), Rollouts (k)=22026.04 | 91.7 | |
| NexAU0Base Model=GPT-5.4 (high reasoning), Rollouts (k)=22026.04 | 87.5 | |
| opencodeBase Model=GPT-5.4 (high reasoning), Rollouts (k)=22026.04 | 75 | |
| terminus-2Base Model=GPT-5.4 (high reasoning), Rollouts (k)=22026.04 | 75 | |
| CodexBase Model=GPT-5.4 (high reasoning), Rollouts (k)=22026.04 | 75 |