-
Notifications
You must be signed in to change notification settings - Fork 3.1k
Pull requests: openai/parameter-golf
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Non-record: 11L s2048 4h on 1xA100 — 1.1104 BPB
#1528
opened Apr 10, 2026 by
xiehuanyi
Loading…
4 of 5 tasks
Record: 6L depth minimalism U-Net sliding window - val_bpb 1.2025
#1527
opened Apr 10, 2026 by
alphastar1111
Loading…
Record: SP8192 + Triple Recurrence + Banking + Fused MLP + Muon 0.97 — val_bpb 1.0778 (3-seed mean)
#1523
opened Apr 10, 2026 by
EthanYangTW
Loading…
Record: SP8192 + Muon 0.97 + 3-Layer Recurrence + Parallel Residuals + TTT — val_bpb 1.0802 (3-seed mean)
#1521
opened Apr 10, 2026 by
aryanbhosale
Contributor
Loading…
SP8192 + Gated Attention + NorMuon + Norm-PCT-Dropout + Legal TTT — val_bpb 1.0824
#1520
opened Apr 10, 2026 by
taka6745
Loading…
5 of 6 tasks
BPB-weighted training loss: align training objective with eval metric
#1519
opened Apr 10, 2026 by
elliottdehn
Loading…
Record: Wider Loop + Per-Pass Embeddings + Tap-In V6 + Legal TTT (1.078825 3-seed mean)
#1518
opened Apr 10, 2026 by
abaybektursun
Contributor
Loading…
Record: Depth Recurrence + Banked Muon + Pre-Quant TTT (18ep) — val_bpb 1.0632 (3-seed mean)
#1517
opened Apr 10, 2026 by
RulinShao
Loading…
Non-Record: Polar Express Muon negative result (1.0805 BPB, +0.0004 vs standard NS5)
#1516
opened Apr 10, 2026 by
dexhunter
Contributor
Loading…
3 tasks done
Non-Record: SP8192 + LeanICQ Compose at Int3 — val_bpb 1.08720 / 15.88 MB
#1515
opened Apr 10, 2026 by
dexhunter
Contributor
Loading…
4 tasks done
Record: SP8192 + Muon 0.97 + Legal Score-First TTT — val_bpb 1.07983 (3-seed mean)
#1514
opened Apr 9, 2026 by
dexhunter
Contributor
Loading…
7 tasks done
Non-record JEPA-style regression transformer submission: VRS (Void Rescue System)
#1513
opened Apr 9, 2026 by
ikermoel
Loading…
Record: Bank QAT + seq4096 + SWA w=256 + QK-Gain 2.5 + PKO — val_bpb 1.1117 (3-seed mean)
#1512
opened Apr 9, 2026 by
Itssshikhar
Loading…
ANS weight compression: 1.6 MB (13.9%) lossless savings over LZMA
#1510
opened Apr 9, 2026 by
OE-GOD
Loading…
Non-record: DepthScale — Parameter-Shared Iterative Transformer (1.1962 BPB)
#1509
opened Apr 9, 2026 by
Lumi-node
Loading…
4 of 5 tasks
Non-record: 11L 3x MLP Seq2048 — val_bpb 1.1791 (8xH100 SXM)
#1505
opened Apr 9, 2026 by
Rohan-Abhilash
Loading…
5 tasks done
Non-record: DP tokenizer beats naive baseline but fails to close gap to SP — 47-run controlled study (1.2206 @ 4096)
#1504
opened Apr 9, 2026 by
Stuckertks09
Loading…
Flow Matching Language Model for Text Generation
#1503
opened Apr 9, 2026 by
vukadinovic936
Loading…
[Non Record] Learn to Learn: Meta-Learning-TTT Redesign — Cross-Chunk FOMAML + Delta-Loss + MetaSGD
#1502
opened Apr 9, 2026 by
SPThole
Loading…
[Non Record] Learn to Learn: Position-Conditional Bigram Hashing + Meta-Learning + TTT Ablation
#1501
opened Apr 9, 2026 by
SPThole
Loading…
SP8192 + Depth Recurrence + Parallel Residuals (14.09MB)
#1499
opened Apr 9, 2026 by
dippatel1994
Loading…
3 tasks
Non-record: Pre-Quant TTT 11ep + Val-Calibrated GPTQ + SLOT-24 — quad-stack synthesis (validation pending compute)
#1498
opened Apr 9, 2026 by
owizdom
Loading…
5 of 7 tasks
Restore non-record submission: 2026-04-08 Vocab1792 FlashMuon LinearScaleInit XSA5LastGated RReLU2 Int6AWQ MixedBits
#1496
opened Apr 9, 2026 by
shram86
Loading…
Add non-record submission: 12L 24min Vocab1792 FlashMuon LinearScaleInit XSA5LastGated RReLU2 Int6AWQ MixedBits
#1495
opened Apr 9, 2026 by
shram86
Loading…
Previous Next
ProTip!
Mix and match filters to narrow down what you’re looking for.