tokenbender

name	pdf-ocr-feedback
description	High-accuracy OCR pipeline using Maj@K consensus voting, structured self-evaluation, and adaptive compute budgets to achieve ≥95% transcription accuracy.

When to Use

Use when transcribing PDF pages via vision model and you need high accuracy — especially for:

• Equations or mathematical notation
• Tables with complex structure (3+ columns, merged cells)

tokenbender

Paper OCR System - Complete Architecture

Overview

A nested agent architecture for high-accuracy PDF OCR and structured note generation.

---

Location