ThesisHuman Model Release Notes

Our most comprehensive academic rewrite, combining a validated natural-text preprocessing stage with the protected fine-tuned rewrite.

• Natural-text Phase 1 with exact-atom and 85%-115% length validation
• Generic atomic ZXQ protection applied once before Phase 2
• Request-level Humanize and Rehumanize history with sanitized process metrics

High-velocity academic humanizer engineered for rapid iterative drafting, concise abstracts, and fast paragraph transformations.

• High-speed generation throughput with 2.5x faster velocity
• Optimized for quick drafting, concise abstracts, and methodology sections
• Strict deterministic formatting and citation preservation

Advanced cross-sentence semantic restructuring and natural cadence variation with zero-hallucination citation locking.

• Erased Claude underlying watermarks and stylistic fingerprints
• Advanced cross-sentence semantic restructuring and natural cadence variation
• Dynamic lexical entropy calibration preserving formal academic tone

Introduced document-level contextual synthesis, enabling the model to analyze the full narrative flow of a manuscript rather than isolated fragments.

• Full document-level contextual awareness eliminating boundary artifacts
• Up to 88.4% human-score reliability on Originality.ai Standard
• Enhanced passive-to-active voice modulation according to IMRAD conventions
• Substantially lowered over-editing on already-humanized introductory sentences

Focused on semantic drift minimization, citation formatting integrity, and scholarly precision in literature reviews.

• Reduced semantic drift by 14% on empirical claim assertions
• Comprehensive parser support for Chicago (Author-Date & Notes-Bib) and IEEE citation styles
• Preservation of numerical findings, p-values, and statistical confidence intervals

Calibrated sentence-length entropy (burstiness) specifically for APA 7th edition research designs and methodology sections.

• Perplexity smoothing on formal academic methodologies and empirical protocols
• Mitigated repetitive transition clichés ('Furthermore', 'Moreover', 'In conclusion')
• Enhanced vocabulary density calibration for peer-reviewed journal criteria

The initial architecture release establishing the proprietary Ghosty academic transformation paradigm.

• Dynamic burstiness balancing across IMRAD manuscript structures
• Unified perplexity smoothing eliminating robotic machine rhythms
• Foundational support for APA and MLA in-text citation masking