Davin/clarity fixes - #147
Merged
Merged
Conversation
added 2 commits
July 29, 2026 01:19
Applied on top of the AAAI-27 checklist commit. Ten are corrections, four
reword sentences that described the paper's own findings as losses. No number
or citation altered (the one numeric delta is a repeated 0.997 supplying a
comparison value that had no referent). Main body is 73 chars shorter.
Corrections:
* "It also keeps its value away from benchmarks of our own design" asserted
the opposite of the intended meaning
* tab:real's caption cited a deterministic-descent column the table does not
contain; now marked "not shown"
* "one seed at 0.227 below the random arm's 0.248" parses as a 0.227
difference rather than an absolute rate
* "far below the ranker, where exact paired McNemar..." had no comparison
value and a dangling "where"
* "outside what this control tests" and "a control it does not run" had no
resolvable antecedent
* "most of it the control returning to its true rate" was unparseable
* "A learned proposal plus refinement is better than a cold start, which
rewards the model..." attached the relative clause to the wrong noun and
asserted as fact what the sentence means to describe as a flawed inference
* "the two classes that never fail" read as the method never failing; it is
the reference solver
* sentence-initial lowercase "supplementary Table S4"
Framing, limited to four sentences:
* the abstract's "Does it improve...? Only narrowly, in a predictable
regime" is now declarative; the facts and the ordering are unchanged
* "The same controls cost us a claim on the value decision"
* "it shrinks our headline claim"
* the Conclusion's "returns a characterization instead"
Deliberately not applied: 21 further framing edits (subsection-title rewrites,
Related Work disclaimer softening, and similar) that fix no error, and two that
collided with the factual corrections in b61aa76, which supersede them.
Known, pre-existing and unchanged by this commit: under the local 2026-style
proxy build the last line of the AI-use statement sits on page 8, which AAAI
reserves for references. Present in c953c3e too; needs confirming against the
real pdfLaTeX build with aaai2027.sty, where line breaking differs.
52/52 number checks. main 8 pages, supplement 4, no unresolved refs or
overfull boxes.
Independent review of all 14 edits found one that altered substance in the
paper's favor, plus three precision issues. All four amended.
* Conclusion had become "those conditions predict that coupling removes the
advantage", converting a measured result into a derivation of our own
theory in the one place a reviewer looks for what was found. Table 1 and
the coupled section report the measurement (no significant win at any of
five dimensions). Now reads "coupling removes the advantage, as those
conditions predict", which keeps the measurement and the confirmation
* "which we characterize exactly" -> "precisely"; "exact" is a technical
term in this paper (exact CAS, exact checker, exact McNemar) and the
characterization is an empirically fitted one-constant law, MAE 0.012
* "controls narrow the value decision" -> "narrow the value-decision
claim"; controls narrow claims, not decisions
* two one-token fixes: a stray comma that loosened which quantity is below
which in the seed-stability parenthetical, and "that line" -> "this
line" where the paragraph already says "this line"
Claim inventory against c953c3e: every disclosed limitation still present at
the same count (CircleLine never solved, 0 of 48 MATH-500 coverage, both
withdrawal disclosures, linear edge not seed-stable, no universal-solver claim,
Jensen-gap caveat, block-decomposability caveat, gains-nothing-at-n=1, the
construction scope limit, the selection-on-noise measurement), and all 24
instances of the scope word "only" retained. Numbers and citations identical.
52/52 number checks. main 8 pages, supplement 4, clean build.
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.
No description provided.