The full battery replicates: 53/58, twice - #43
Merged
Conversation
An independent 62-question re-run with the derivative field live: arithmetic 18/20 -> 18/20 logic 8/8 -> 8/8 units 8/8 -> 8/8 calculus 5/8 -> 6/8 physics 14/14 -> 13/14 totals 53/58 -> 53/58 Identical totals across two full runs -- after a session in which identical configurations at n=18 swung 28 points, two matching 58-question totals is what stability actually looks like, and it is pinned as the replication that makes the nine-in-ten claim worth trusting. The derivative field: zero uses again, now confirmed across a full run. qwen3 applies the chain rule itself and hands the arithmetic to calc; the field waits for the model that reaches for it. The persistent misses are stable too -- division_with_remainder and comparison miss in both runs -- which marks them as systematic rather than noise, and as the next thing worth a mechanism. No code changes; one claims file and one pin.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
An independent 62-question re-run with the derivative field live.
Identical totals across two full runs. After a session in which identical configurations at n=18 swung 28 points, two matching 58-question totals is what stability actually looks like — pinned as the replication that makes the nine-in-ten claim worth trusting.
The derivative field: zero uses again, now confirmed across a full run — qwen3 applies the chain rule itself and hands the arithmetic to
calc. The persistent misses are stable too:division_with_remainderandcomparisonmiss in both runs, marking them as systematic rather than noise — and as the next thing worth a mechanism.No code changes; one claims file and one pin.