Update README leaderboard for April records

Author: cocohearts

README.md

@@ -30,6 +30,14 @@ Happy training!
 
 | Run | Score | Author | Summary | Date | Info |
 |-----|------:|--------|---------|------|------|
+| SP8192 + 3-Layer Recurrence + Parallel Residuals + Legal TTT | 1.0810 | bigbag | On PR #1493 (submitted 2026-04-09 07:18 UTC): 3-layer recurrence over layers 3-5, earlier recurrence start, QK-Gain 5.25, tuned WD/LR/EMA, and legal score-first SGD TTT on the PR #1477/#1394 stack | 2026-04-09 | [info](records/track_10min_16mb/2026-04-09_SP8192_3LayerRecur_ParResid_QK525_LegalTTT/README.md) |
+| SP8192 + Parallel Residuals + Score-First TTT | 1.0822 | aryanbhosale | On PR #1477 (submitted 2026-04-08 17:12 UTC): added layer-7+ parallel residual lanes to the PR #1413 SP8192 + legal score-first TTT stack | 2026-04-08 | [info](records/track_10min_16mb/2026-04-08_SP8192_ParallelResid_ScoreFirstTTT/README.md) |
+| SP8192 + QK-Gain 5 + Legal Score-First TTT | 1.0828 | dexhunter | On PR #1413 (submitted 2026-04-06 11:28 UTC): raised QK-Gain from 4.0 to 5.0 and added chunked score-first TTT after scoring each chunk on the PR #1394 SP8192 stack | 2026-04-06 | [info](records/track_10min_16mb/2026-04-06_SP8192_QK5_LegalTTT_1.0828/README.md) |
+| SP8192 + GPTQ Embeddings + Depth Recurrence + SDClip | 1.0856 | Kevin Clark | On PR #1394 (submitted 2026-04-05 21:07 UTC): moved from SP4096 to SP8192, GPTQ-quantized embeddings, looped layers 4-5 twice, used MuonEq-R, and replaced clip search with std-based SDClip | 2026-04-05 | [info](records/track_10min_16mb/2026-04-05_SP8192_GPTQ-Embeddings_SDClip_Loop45x2/README.md) |
+| SP4096 + Depth Recurrence + Parallel Residuals + MuonEq-R | 1.0897 | aryanbhosale | On PR #1334 (submitted 2026-04-04 09:33 UTC): combined SP4096, layers 4-5 depth recurrence, layer-7+ parallel residual lanes, MuonEq-R, QK-Gain 5.0, GPTQ int6, and no eval-time adaptation | 2026-04-04 | [info](records/track_10min_16mb/2026-04-04_SP4096_DepthRecurrence_ParallelResid_MuonEqR/README.md) |
+| MuonEq-R + Depth Recurrence + WD=0.090 + All-Int6 GPTQ | 1.0912 | dexhunter | On PR #1285 (submitted 2026-04-03 05:34 UTC): added MuonEq-R, repeated layers 4-5, raised weight decay to 0.090 for Brotli headroom, and kept all 66 GPTQ matrix layers at int6 | 2026-04-03 | [info](records/track_10min_16mb/2026-04-03_MuonEqR_DepthRecurrence_WD090_AllInt6/README.md) |
+| 4096-Vocab + Larger Model + High WD + Simplifications | 1.0979 | Kevin Clark | On PR #1218 (submitted 2026-04-01 11:55 UTC): switched to SP4096, widened MLP to 4x, increased weight decay, added GPTQ/Brotli, and removed TTT, hash embeddings, SmearGate, value residuals, and other auxiliaries | 2026-04-01 | [info](records/track_10min_16mb/2026-04-01_Vocab4096_MLPMult4_WD085/README.md) |
+| Parallel Residuals + Mini Depth Recurrence | 1.1063 | Marko Sisovic | On PR #1204 (submitted 2026-04-01 00:46 UTC): added delayed mini recurrence on layers 4-5, untied repeated MLPs, layer-7+ parallel attention/MLP residual lanes, and AR self-generated GPTQ calibration | 2026-03-31 | [info](records/track_10min_16mb/2026-03-31_ParallelResiduals_MiniDepthRecurrence/README.md) |
 | 11L AR Self-Gen GPTQ + XSA | 1.1147 | abaybektursun | On PR #1019: Self-Generated GPTQ Calibration Data + all-layer XSA on the PR #549 stack | 2026-03-25 | [info](records/track_10min_16mb/2026-03-25_ValCalib_GPTQ_XSA_BigramHash3072/README.md) |
 | LeakyReLU² + Legal Score-First TTT + Parallel Muon | 1.1194 | abaybektursun | On PR #549: LeakyReLU(0.5)^2 + TTT + Parallel Muon on the PR #414 stack | 2026-03-23 | [info](records/track_10min_16mb/2026-03-23_LeakyReLU_LegalTTT_ParallelMuon/README.md) |
 | 11L EMA + GPTQ-lite + warmdown3500 | 1.1228 | signalrush | On PR #374: GPTQ-lite clip search + EMA, plus warmdown3500 and QAT@0.15 | 2026-03-22 | [info](records/track_10min_16mb/2026-03-22_11L_EMA_GPTQ-lite_warmdown3500_QAT015_1.1233/README.md) |