35%
Fidelity and completeness
Preserve meaning and deliberate ambiguity. No invented, omitted or reversed ideas.
LitMundi · v2 · ≥8.6/10
AI review. Open reading. Ongoing correction.
Under the current policy, a complete translation may be published after an AI literary review scores at least 8.6/10 and the rights and integrity checks pass. Human readers can then review it, report errors and suggest improvements.
The literary standard
25%
Voice, rhythm and register
Keep the author’s tone, repetition, cadence and stylistic choices.
15%
Naturalness
Write idiomatic literary prose that withstands reading aloud.
15%
Imagery and terminology
Preserve metaphors, cultural references and recurring terms.
10%
Editorial mechanics
Use accurate grammar, spelling, punctuation and typography.
Requirements for publication
- At least one current AI literary review with a weighted score ≥8.6/10, covering every canonical passage.
- Complete original and translation texts with explicit passage alignment.
- No unresolved major or critical translation errors.
- Review evidence bound to the exact source and translation SHA-256 hashes; an old score cannot approve changed text.
- Documented source provenance, lawful publication rights and applicable open-content permissions.
- Clear AI-review labeling and a public correction channel. Human review and reader feedback can follow publication.
AI-reviewed editions are labeled as AI-reviewed. They do not imply independent human approval.
An AI score is an editorial judgment, not an objective guarantee of literary quality. The review record, exact text version, sources and permissions remain visible. Corrections can lead to a new review, revision or withdrawal.
Earlier policy and review history
Version 1 required two independent qualified human reviews scoring at least 9.0/10. That policy and its records are retained as history; it is not the prerequisite for AI-reviewed publication under version 2.