host=timan1.cs.illinois.edu gpu=0 start=2026-06-16T10:23:58-05:00
ep20 val_mae=[0.316, 0.393]
ep40 val_mae=[0.357, 0.365]
ep60 val_mae=[0.165, 0.247]
ep80 val_mae=[0.171, 0.226]
ep100 val_mae=[0.415, 0.502]
ep120 val_mae=[0.183, 0.27]
ep140 val_mae=[0.149, 0.203]
ep160 val_mae=[0.172, 0.243]
ep180 val_mae=[0.171, 0.205]
ep200 val_mae=[0.176, 0.212]
[rec_full_sig0.0_K1_bestq_T3_ns3_s0] test_mae=[0.145, 0.198] oracle@K=[0.145, 0.198] @ep140 (209.3s)
  wrote /home/yurenh2/rrog/runs/ckpt_rec_full_sig0.0_K1_bestq_T3_ns3_s0.pt
ep20 val_mae=[0.353, 0.393]
ep40 val_mae=[0.281, 0.337]
ep60 val_mae=[0.217, 0.308]
ep80 val_mae=[0.194, 0.285]
ep100 val_mae=[0.185, 0.313]
ep120 val_mae=[0.21, 0.373]
ep140 val_mae=[0.531, 0.684]
ep160 val_mae=[0.471, 0.599]
ep180 val_mae=[0.56, 0.6]
ep200 val_mae=[0.595, 0.645]
[rec_1step_sig0.0_K1_bestq_T3_ns3_s0] test_mae=[0.211, 0.293] oracle@K=[0.211, 0.293] @ep80 (168.0s)
  wrote /home/yurenh2/rrog/runs/ckpt_rec_1step_sig0.0_K1_bestq_T3_ns3_s0.pt
===== LE diagnostic (lambda1: success vs failure) =====
[full] n=300 fail_rate=0.15 | lambda1 SUCC mean -0.0402 std 0.1253 (n=256) | FAIL mean +0.0156 std 0.1058 (n=44) | sep(fail-succ)=+0.0558 | AUROC(fail|lambda1)=0.633 | mean_lambda1=-0.0320
[1step] n=300 fail_rate=0.29 | lambda1 SUCC mean -0.2247 std 0.1226 (n=213) | FAIL mean -0.1706 std 0.1229 (n=87) | sep(fail-succ)=+0.0542 | AUROC(fail|lambda1)=0.630 | mean_lambda1=-0.2090
done=2026-06-16T10:31:15-05:00