Record: MuonEq-R + Depth Recurrence + WD=0.090 + All-Int6 GPTQ — val_bpb 1.0912 (3-seed mean)#1285
Merged
cocohearts merged 1 commit intoopenai:mainfrom Apr 9, 2026
Conversation
….0912 (3-seed mean) WD-quantization synergy: higher weight decay (0.090 vs 0.085) compresses 5% better, creating headroom for ALL 66 layers at int6 precision. The extra quantization quality more than recovers the WD BPP cost. 3-seed mean: 1.0912 BPB / 2.5106 nats (seeds 42, 0, 1337) All seeds under 16MB with 32K+ margins. No TTT, no SLOT, no eval-time adaptation. Built on PR openai#1218 by @clarkkev. Improves PR openai#1260 (1.0929) by 0.0017 BPP.
This was referenced Apr 3, 2026
dexhunter
added a commit
to dexhunter/parameter-golf
that referenced
this pull request
Apr 4, 2026
…b 1.0900 (3-seed mean) 3-layer depth recurrence (layers 3,4,5) with WD-LR synergy: higher WD (0.095) compresses for all-int6 headroom, higher MLR (0.022) recovers quality. All 66 layers at int6 precision. 3-seed mean: 1.0900 BPP / 2.5077 nats (seeds 42, 0, 7) All seeds under 16MB with 36K+ margins. No TTT, no SLOT, no eval-time adaptation. Improves PR openai#1285 (1.0912) by 0.0013 BPP. Beats PR openai#1218 by 0.0079. Built on PR openai#1218 by @clarkkev.
chandra447
added a commit
to chandra447/parameter-golf
that referenced
this pull request
Apr 4, 2026
Hypothesis: Polar Express 4-step minimax NS on top of full PR openai#1334 stack Expected delta: ~-0.001 to -0.002 BPB from 1.0897 baseline Key changes vs PR openai#1334: - Polar Express Newton-Schulz (4-step minimax coefficients, arXiv:2505.16932) - MATRIX_LR=0.022 (validated for WD=0.090) - MUON_WD=0.090 (PR openai#1285/1334 optimal for 2-layer recurrence) - NoPE explicitly disabled (nope_every_n=0) after critique - Trackio experiment tracking added Stack: SP4096 vocab + MLP 4x + WD=0.090 + MuonEq-R + QK-Gain 5.0 + Depth recurrence L4-5 (step 3000) + Parallel residuals L7+ + Brotli
EthanYangTW
added a commit
to EthanYangTW/parameter-golf
that referenced
this pull request
Apr 4, 2026
…el spectral test-time training Base: clarkkev openai#1218 (1.0974 BPB, 4096 vocab, brotli, 34M params) Added: depth recurrence L4,5 (from openai#1285), MuonEq-R, WD=0.09 Novel: Spectral TTT — adapt singular values at eval time (8192 params) Target: ~1.085 BPB
AnubhavBharadwaaj
added a commit
to AnubhavBharadwaaj/parameter-golf
that referenced
this pull request
Apr 6, 2026
SH-Tan
pushed a commit
to SH-Tan/parameter-golf
that referenced
this pull request
Apr 9, 2026
…0-allint6 Record: MuonEq-R + Depth Recurrence + WD=0.090 + All-Int6 GPTQ — val_bpb 1.0912 (3-seed mean)
sunnypatneedi
pushed a commit
to sunnypatneedi/parameter-golf
that referenced
this pull request
Apr 12, 2026
…1.01710 Merged SOTA changed from 1.1147 to 1.0810 (PR openai#1493, bigbag, 2026-04-09). Six PRs merged in 5 days (PRs openai#1334, openai#1285, openai#1394, openai#1412, openai#1413, openai#1477, openai#1493). New target: ≤1.0760 val_bpb. 18 days to deadline. Key findings: - GDN-Hybrid (PR openai#1564): 1.01710 BPB, no TTT/SLOT — monitor for organizer review - VarLen Attention + Doc-TTT (PR openai#1560): 1.07406 BPB — implement next - TMA Megakernel + Tap-In (PR openai#1555): 1.07636 BPB — add after openai#1560 - PR openai#731 n-gram (dense count + Laplace): reviewer says LOOKS CLEAN, awaiting 3rd seed - PR openai#758: major legality flags, do not implement Updated CLAUDE.md: Competition Strategy, Technique Reference, Lessons Learned (Session 9). Updated logs/daily_research.md: new 2026-04-12 entry prepended. https://claude.ai/code/session_011WyxjcwdigLhMFQDjLL5ss
tns15june
pushed a commit
to tns15june/parameter-golf
that referenced
this pull request
Apr 19, 2026
Frontier records (PR openai#1285 MuonEq-R + WD=0.090, PR openai#1218 WD=0.085) use AdamW-style decoupled weight decay on the Muon optimizer. Add the knob with default 0.0 (backward-compatible). Applied as p.data.mul_(1 - lr * wd) before the Muon matrix update. MuonEq-R (row-normalized) variant is not ported — it would need more line budget than we have on this branch. WD alone accounts for the majority of that record's improvement per the commit notes. dev/run_frontier.sh sets MUON_WEIGHT_DECAY=0.09 by default. Also inlined restore_low_dim_params_to_fp32 at its single call site to free lines for this change. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
8 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Key Innovation: WD-Quantization Synergy
Higher WD (0.090 vs 0.085) → smaller weights → 5% better brotli compression → enough headroom for ALL 66 layers at int6 precision. The quantization quality gain exceeds the WD BPP cost:
Results (8xH100 80GB SXM, PyTorch 2.9.1+cu128)
Changes from PR #1218
Credits
Test plan