Skip to content

[proof search] search-algorithm.tex, tableaux.tex: G3c search and Tc tableau search - #441

Open
KokunoYumeto wants to merge 7 commits into
OpenLogicProject:masterfrom
KokunoYumeto:translation-findings-proof-search
Open

KokunoYumeto wants to merge 7 commits into
OpenLogicProject:masterfrom
KokunoYumeto:translation-findings-proof-search

Conversation

@KokunoYumeto

Copy link
Copy Markdown
Contributor

While translating the Open Logic Project into eighteen languages, the AI checks of the translations kept finding small errors in the English source. We reported them as issues (#432, #433, #435, #436), and you have been integrating them by hand. These pull requests may save you that work: each fix is a separate commit, checked against master a216aa64. With them, open-logic-complete builds as on master (no errors, no new undefined references), and the changed proof-theory sections, which that driver does not include, compile on their own as on master. Merge any, all or none; each branch starts from master and merges on its own.

  • search-algorithm.tex: stop only at atomic A on both sides, since G3c's axioms require C atomic.
  • search-algorithm.tex: new formulas get indices k+1 (and k+2), since k is the maximal index already in use; the right-∃ rule is labelled \RightR{\lexists}.
  • search-algorithm.tex: the fallback term is the first constant in C, not c₀, so that C keeps containing every constant of the sequent (the eigenvariable argument needs this).
  • tableaux.tex: a branch containing T⊥ is closed (Tc has no rule for it; G3c has ⊥ on the left as an axiom).
  • tableaux.tex: "tableau is closed", "closed tableau" (next to the previous fix).
  • tableaux.tex: the search picks the first unchecked non-atomic formula other than T∀/F∃ (atomic ones are never checked off).
  • tableaux.tex: a finite open branch: T∀/F∃ formulas are never checked off; they have been applied to every available term.

Prepared by AI: the fixes were written by a Claude agent (Claude Opus 5.5 at effort medium), and the mathematical ones were checked by Claude Opus 5.5 at effort max.

🤖 Generated with Claude Code

KokunoYumeto and others added 7 commits November 15, 2026 13:32
… require

G3c axioms have atomic C; a compound A on both sides is not an axiom, so erasing indices gives no G3c proof.

Ledger: OLI-PT-SEARCH-ATOMIC-CLOSURE, OLCMP-191, BN-SRC-920

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…ight-exists label

k is the maximal index already in use, so index k repeats one; the rule shown with the existential on the right is the right rule.

Ledger: BN-SRC-600, OLCMP-192, BN-SRC-599, BN-SRC-601:rule-side-component, OLCMP-193

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…eps covering the sequent

c_0 need not be in C; using it breaks the invariant on which the eigenvariable condition rests.

Ledger: OLI-PT-SEARCH-FALLBACK-CONSTANT, BN-SRC-888, OLCMP-194

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Tc has no rule for T falsum, and the search algorithm already treats falsum on the left as an axiom.

Ledger: BN-SRC-889, OLCMP-198

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Typos (kept in this bundle because they are next to the falsum fix).

Ledger: (found 15 Nov)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Atomic signed formulas have no rule and are never checked off, so the first unchecked formula can block the search.

Ledger: BN-SRC-892, OLCMP-196

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…t saturation means for them

Those formulas are never checked off by the text's own rule, so the finite case as stated cannot occur.

Ledger: BN-SRC-890, OLCMP-197

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant