A 2B-parameter multimodal model fine-tuned with QLoRA on a gaming PC beat larger OCR baselines on historical Spanish document transcription, while Tesseract fine-tuning offered no gain.
Sánchez-Salido, A
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Transcribing Spanish Texts from the Past: Experiments with Transkribus, Tesseract and Granite
A 2B-parameter multimodal model fine-tuned with QLoRA on a gaming PC beat larger OCR baselines on historical Spanish document transcription, while Tesseract fine-tuning offered no gain.