Merge pull request #561 from openai/codex/update-readme-leaderboard-merged-records - #1

Open

peterpodj wants to merge 79 commits into

0hq-patch-1from

Commits on Mar 18, 2026

Merge pull request #35 from openai/0hq-patch-1
0hq
authored
0f9518a
Update train_gpt.py
0hq
authored
886cc5b
Update train_gpt_mlx.py
0hq
authored
de13248
Update README.md
0hq
authored
954a158

Commits on Mar 19, 2026

Record: Seq4096 + Sliding Window Eval, val_bpb=1.1808
aquariouseworkman
committed
9d318e7
Non-record: SwiGLU + warmdown fix + quarter batch (1x5090, 1.3281 bpb)
NishantDahal
committed
b8a1426
Record: Mixed Quant (int6+int8) + Sliding Window, val_bpb=1.1630
aquariouseworkman
committed
f85c837
Submission: 10L + lower LR + fp16 embed + int6 middle layers (val_bpb 1.2129)

Arun Iyer
and
Copilot
committed
73a4a9d
Add MLX_EAGER_EVAL flag to further reduce memory pressure by force-evaluating the graph after each sub-batch step
sandsevenone
committed
6a08c9d
Merge pull request #100 from sandsevenone/mlx_eager_eval
0hq
authored
2081ba1
fp16 tied embedding + lr/warmdown tuning — val_bpb 1.2197 (#42)

chonchiog
and
Copilot
authored
a5eb9ed
Update README.md (#105)
0hq
authored
b5ea566
clarify torch version
cocohearts
committed
6e3e90d
SOTA attempt (val_bpb=1.2064) (#49)

spokane-way
and
spokane-way
authored
e89fcf8
Update README.md
0hq
authored
194bb87
Add record: Sliding Window Eval (stride=64), val_bpb=1.1925 (#50)
mattqlf
authored
d84a3e8
Update submission: MLP 3x + QAT + Int6 + Sliding Window (val_bpb 1.1652)

Arun Iyer
and
Copilot
committed
8e4f5d1
Fix: score final partial window in sliding window eval (#124)
mattqlf
authored
3a6fec7
Update README.md
0hq
authored
6b40978
New SOTA attempt (#52)

spokane-way
and
spokane-way
authored
78c24e2
Update README.md
0hq
authored
b87b883
Update README.md
0hq
authored
2d6e9e0
Update README.md
0hq
authored
ce6cf9a
Record: Sliding Window + FP16 Embed + 10L + Muon WD + Overtone Init (val_bpb=1.1748) (#60)

notapplica
and
claude
authored
9fbdf8c
Update README.md
0hq
authored
ad7b62c
Int6 + MLP 3x + sliding window: val_bpb=1.1574 (#61)
saml212
authored
555669e
Update README.md
0hq
authored
cfa5726
Update README.md
0hq
authored
f3897c1
Update README.md
0hq
authored
5353524
Update README.md
0hq
authored
d2bd760
Add Seq2048 + FP16 Tied Embedding submission (mean val_bpb 1.2067)

yahya010
and
claude
committed
34fccfb
Update submission: 10L + int6 mid + sliding window (mean val_bpb 1.1793)

yahya010
and
claude
committed
c2b3621
Update: full int6+zstd, MLP 1344, Muon 0.99 (mean val_bpb 1.1632)

yahya010
and
claude
committed
9e7a4b8
Update: STE int6 QAT, zero quant gap (mean val_bpb 1.1598)

yahya010
and
claude
committed
510e3f6
Update README.md
0hq
authored
ae88208
Record: 10L Mixed Precision: val_bpb=1.2147 (10 layers + int6 middle layers) (#39)
nanlliu
authored
9ac12c2
commit ttt record (#77)
samacqua
authored
bd2463a
Update README.md
0hq
authored
5e29bfd
Update README.md
cocohearts
authored
45bbccf

Commits on Mar 20, 2026

SmearGate + OrthoInit + Muon WD + Int6 STE QAT + MLP 3x + Sliding Window
aquariouseworkman
committed
e109841
Merge branch 'openai:main' into main
aquariouseworkman
authored
3aface5
Add submission: 2026-03-20_Int6_MLP3x_SmearGate_BigramHash_MuonWD_SWA (mean val_bpb=1.1483, 3 seeds)
Raahil Shah
committed
14cdf6f
Update submission: 2026-03-20_Int6_MLP3x_SmearGate_BigramHash_MuonWD_SWA — improved config (Muon WD=0.04, SWA every 50), mean val_bpb=1.1458
Raahil Shah
committed
3c458dc
Record: 10L Int5-MLP + MuonWD=0.04 + SWA/50 (val_bpb=1.1453)
thwu1
committed
ab9ecb2
Update records/track_10min_16mb/2026-03-20_10L_Int5MLP_MuonWD04_SWA50/submission.json

thwu1
and
Copilot
authored
a320f7c
Update: 11L MLP3x + WD=0.04 + zstd-22 (val_bpb 1.1502)

Arun Iyer
and
Copilot
committed
1641a23
Update: val_bpb=1.1428 (mean 3 seeds) — bigram=10240 + SWA(0.4) + WD=0.04
Tianhao Wu
committed
9f9d533
Add 3-seed training logs (seed=42, 1337, 2024)
Tianhao Wu
committed
1a8be36
Merge pull request #63 from yahya010/submission/seq2048-fp16emb
cocohearts
authored
80f7a21
Merge pull request #65 from aquariouseworkman/main
cocohearts
authored
5b26c56
Merge pull request #73 from NishantDahal/swiglu-warmdown-1x5090
cocohearts
authored
e821922
Merge pull request #86 from aruniyer/submission/10L-lowlr-fp16embed-int6
cocohearts
authored
b774930
Merge pull request #162 from raahilshah/submission/2026-03-20_Int6_MLP3x_SmearGate_BigramHash_MuonWD_SWA
cocohearts
authored
8b2b17e
Merge pull request #180 from thwu1/10L-int5mlp-wd04-swa50
cocohearts
authored
ee82226
Update README.md
valerio-oai
authored
4958197
Update README.md
valerio-oai
authored
f427d59
Create leaderboard-best-score-over-time.svg
valerio-oai
authored
f6e75fe
Update README.md
valerio-oai
authored
a9d5c77
Delete assets directory
valerio-oai
authored
e721424
Record: 11L + Efficient Partial XSA (val_bpb: 1.1307) — NEW SOTA
unnir
committed
a81f85b
Merge pull request #255 from openai/valerio-oai-patch-1
cocohearts
authored
2d7199d
Restore train_gpt.py before bd2463a

yuzhougu-oai
and
codex
committed
48404cd
Merge pull request #269 from openai/revert-train-gpt-pre-bd2463a
cocohearts
authored
3c1f8b3
Update README.md
cocohearts
authored
2cff08e
Update README.md
0hq
authored
9a60d16
Record: 11L XSA + EMA + Int6 MLP3x + WD=0.04 (val_bpb: 1.1271)
jfprincz
committed
d758328
Update README.md
0hq
authored
fc6332a

Commits on Mar 21, 2026

Fix grammar in README
sha-huang
authored
8926429
Merge pull request #350 from sha-huang/hs-patch-grammar
cocohearts
authored
0f51451
Record: 11L Partial RoPE + LN Scale + EMA + XSA4 (val_bpb: 1.1248)
jfprincz
committed
2951651

Commits on Mar 22, 2026

Record: 11L EMA + GPTQ-lite + warmdown3500 + QAT@0.15 (val_bpb=1.1233, 3-seed mean)
signalrush
committed
15776db

Commits on Mar 23, 2026