Record: SP4096 + Compressibility Regularization — val_bpb 1.11349 (6-seed mean)#1508
Closed
jpfeiffe wants to merge 2 commits intoopenai:mainfrom
Closed
Record: SP4096 + Compressibility Regularization — val_bpb 1.11349 (6-seed mean)#1508jpfeiffe wants to merge 2 commits intoopenai:mainfrom
jpfeiffe wants to merge 2 commits intoopenai:mainfrom
Conversation
…seed mean) SP4096 tokenizer + WARMDOWN_WD_MULT=2.0 + Brotli-11 compression on the openai#1019 GPTQ+XSA base. 6-seed mean 1.11349 vs openai#1019's 1.11474. Welch t=-4.19, df=6.6, p=0.00289 (one-sided). No TTT, no eval-time compute. WD compression frees ~1.5MB which absorbs SP4096's +1.1MB embedding cost. Pruning 0% on all 6 seeds. Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
- Removed all references to other PRs (openai#365, openai#1089, etc.) - Removed "novel" language - Brotli: "selected when smaller than LZMA-9" (matches code) - WD ramp: accurately describes group["weight_decay"] mechanism - Added Verification section pointing to support files - Streamlined Changes section Co-Authored-By: Claude Opus 4.6 (1M context) <[email protected]>
Author
|
Other work has been merged in since filed. |
| @@ -0,0 +1,28 @@ | |||
| # BPB Verification | |||
There was a problem hiding this comment.
Suggested change
| # BPB Verification | |
| # BPB Verification[]() |
| @@ -0,0 +1,34 @@ | |||
| # Data Lineage Verification | |||
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
6-seed mean SW val_bpb: 1.11349
#1019 3-seed mean: 1.11473
Delta: -0.00125
Welch's t-test (one-sided): t=-4.19, df≈6.6, p=0.00289
Three changes from #1019:
docs_sha256 verified identical to willdepueoai/parameter-golf)
than lzma (true for all 6 seeds)
All artifacts under 16MB. Pruning: 0% across all 6 seeds.
See README for full mechanism, statistical details, and
reproduction. bpb_verification.md and data_lineage.md cover
the manual BPB recomputation and data lineage proof.
Files:
records/track_10min_16mb/2026-04-08_SP4096_WD2_brotli_jpfeiffe/