Qwen3-VL-4B-Trithemius

Qwen3-VL-4B Trithemius OCR

Qwen3-VL-4B Trithemius is a specialized vision-language model for reading and transcribing early printed works by Johannes Trithemius (1462–1516). Built on Qwen3-VL-4B-Instruct, it was fine-tuned using page-image and transcription pairs derived from public-domain scans held by libraries and other cultural institutions. The model supports the OCR and edition-building work of the open Trithemius Corpus.

This repository contains the full merged model in BF16 safetensors format for use with Transformers and for downstream conversion. For a smaller llama.cpp release, see Carlosian/trithemius-q6-models.

235B teacher distillation

Training used OCR labels generated or adjudicated by Qwen3-VL-235B-A22B-Instruct as a teacher for the 4B student. In the project's page-level OCR comparison, the resulting 4B Trithemius model matched the 235B teacher on 86% of evaluated pages.

This is a project-specific OCR agreement rate—not character error rate, not a claim of general vision-language parity, and not evidence that the 4B model matches the 235B teacher outside this historical-OCR task.

Files

The model weights are split across two safetensors shards and mapped by model.safetensors.index.json. The repository also includes the model configuration, processor configuration, tokenizer, chat template, and generation configuration required for inference.

Recommended use

This is an OCR/transcription model, not an English translation model. Supply one page image at a time and request transcription rather than commentary, modernization, or reconstruction.

Example instruction:

Transcribe this historical printed page faithfully. Preserve line and paragraph
structure, headings, names, dates, abbreviations, and visible marginal or
tabular material. Do not translate or silently repair uncertain text. Mark text
that cannot be read reliably as [unclear]. Return only the transcription.

For dense pages, splitting the image into overlapping regions is safer than simply increasing the generation limit. A larger token cap can extend a repetition loop instead of recovering missing text.

What shipped from this model

The model supplied the Latin OCR witness for 27 Trithemius 4B editions published in the Trithemius Corpus. Those are the project's recommended English reading editions where available. Their English was produced separately through a translation, assessment, and full-work remediation pipeline.

The model card deliberately separates OCR quality from downstream translation quality. Direct inspection of 3,902 pages from eight difficult works found body-page OCR to be generally clean, with total degenerate-page rates of approximately 0.2–3.3% per audited work. The larger historical failure cluster came from an unguarded downstream translation run that could drift to neighboring pages; that translation failure is not a benchmark for this OCR model.

See the corpus Methodology for the complete OCR, translation, grading, and remediation record.

Strengths

• Early printed Latin in blackletter and worn roman type
• Long-s and common ligature handling
• Preservation of headings, lineation, lists, and page structure
• Recovery of names and abbreviated forms on the target material
• Conservative transcription with [unclear] when prompted

Limitations

• This is a domain fine-tune, not a general-purpose OCR benchmark winner.
• Dense catalogues, cipher alphabets, tables, Greek, Hebrew, marginalia, stamps, and severely damaged leaves remain difficult.
• The model can repeat lines or enumerative content, especially on degraded or token-capped pages.
• It can generate plausible-looking letter sequences where the page is unreadable. Compare consequential readings with the facsimile.
• The published T4B translations include downstream translation and editorial work and are not raw demonstrations of model-only output.

Training data and provenance

Training examples were derived from public-domain institutional page scans and transcription pairs assembled during the Trithemius Corpus project. Source code, methodology, work metadata, provenance records, and published OCR witnesses are available in the Trithemius Corpus repository.

The released model should not be taken to imply that libraries or scan providers endorsed the fine-tune. Source institutions retain attribution for their digitized holdings; the underlying historical works are public domain.

License

Released under the Apache License 2.0, consistent with the Qwen base model. Users remain responsible for complying with licenses or terms attached to images they process.

Citation

Fabin, Ian Carlos (2026). Qwen3-VL-4B Trithemius OCR. Hugging Face model repository.
https://huggingface.co/Carlosian/Qwen3-VL-4B-Trithemius

Author

Developed and assembled by Ian Carlos Fabin (Carlosian).