Fix submission.json: val_bpb 1.1219→1.1220, correct artifact size

Author: abaybektursun

records/track_10min_16mb/2026-03-22_LegalTTT_ParallelMuon_1.1213/submission.json

@@ -1,9 +1,9 @@
 {
-  "name": "Legal Score-First TTT + Parallel Muon + Parameter Banking (val_bpb=1.1219, 3-seed mean)",
-  "val_loss": 1.8942,
-  "val_bpb": 1.1219,
-  "bytes_total": 15841722,
-  "blurb": "Legal score-first TTT (PR #461 recipe: SGD+momentum, 3 epochs/chunk, freeze 2 blocks, 32K chunks) on #414 stack with Parameter Banking + Parallel Muon (PR #399). Every token scored BEFORE model adapts, enforced by inference_mode. 3-seed mean: pre-TTT 1.1239, post-TTT 1.1219 (std 0.0005). TTT eval: ~400s. Built on PR #414 by @signalrush, TTT from PR #461 by @anantdgoel, optimizer from PR #399 by @abaybektursun.",
+  "name": "Legal Score-First TTT + Parallel Muon + Parameter Banking (val_bpb=1.1220, 3-seed mean)",
+  "val_loss": 1.8944,
+  "val_bpb": 1.1220,
+  "bytes_total": 15972608,
+  "blurb": "Legal score-first TTT (PR #461 recipe: SGD+momentum, 3 epochs/chunk, freeze 2 blocks, 32K chunks) on #414 stack with Parameter Banking + Parallel Muon (PR #399). Every token scored BEFORE model adapts, enforced by inference_mode. 3-seed mean: pre-TTT 1.1239, post-TTT 1.1220 (std 0.0003). TTT eval: ~400s. Built on PR #414 by @signalrush, TTT from PR #461 by @anantdgoel, optimizer from PR #399 by @abaybektursun.",
   "author": "abaybektursun",
   "github_id": "abaybektursun",
   "date": "2026-03-22"