docs(EXPERIMENTS): E5 verdict — gate FAILED at 5.0853, not adopted - #129
Merged
Conversation
MTP-aux (beta 0.15, 3 steps) cost +0.26pp vs the 4.8218 control; prediction 4.5-4.75 missed. Teacher reproduces 2.289 (control intact). Confound disclosed: preemption at ~8,650 + resume from a checkpoint whose head write was interrupted — final 23% of steps trained with a re-initialized head, so the delta cannot be cleanly split between MTP-aux and the perturbation. Negative result logged; nothing shipped.
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Honest negative: MTP-aux at beta 0.15 / 3 steps scored 5.0853 full-set windowed DER vs the 4.5218 gate and the 4.8218 control (+0.26pp); registered prediction 4.5-4.75 missed. Teacher reproduces 2.289 — harness intact. Confound fully disclosed: the run was preempted at ~8,650 and resumed from a checkpoint whose head write was interrupted, so the final 23 percent of steps trained with a re-initialized aux head — the delta cannot be cleanly split between MTP-aux and the perturbation, and a clean rerun is the optional follow-up. Nothing shipped; the shipped artifact stays vanilla. E6 (Tashkeela mix) unaffected and still training.