phase1_eval_results.txt

=== TIL-26-AE Phase 1 Evaluation Results ===
Model: phase1_final.zip (500k steps)
Episodes: 100
Win Rate: 92.0% (92/100)
Avg Reward: 180.1
Avg Episode Length: 200.0
Avg Bombs/Episode: 20.4
Survival Rate (198+ steps): 100.0%