| { | |
| "model_stage": "grpo", | |
| "policy": "hf", | |
| "model": "Qwen/Qwen2.5-7B-Instruct", | |
| "adapter": "outputs/grpo_adapter", | |
| "tasks": [ | |
| "e1_launch_readiness", | |
| "m1_budget_reallocation", | |
| "h1_acquisition_defence" | |
| ], | |
| "episodes": 1, | |
| "max_steps": 30, | |
| "max_new_tokens": 1536, | |
| "eval_file": "results/grpo_eval.jsonl", | |
| "metrics_by_task": { | |
| "e1_launch_readiness": { | |
| "avg_terminal_reward": 0.91, | |
| "avg_verifier_pass_rate": 1.0, | |
| "success_rate": 1.0 | |
| }, | |
| "m1_budget_reallocation": { | |
| "avg_terminal_reward": 0.943333, | |
| "avg_verifier_pass_rate": 1.0, | |
| "success_rate": 1.0 | |
| }, | |
| "h1_acquisition_defence": { | |
| "avg_terminal_reward": 0.6825, | |
| "avg_verifier_pass_rate": 0.75, | |
| "success_rate": 0.0 | |
| } | |
| } | |
| } |