PolyOCR-Venus: Unified OCR Foundation Models for Text-Centric Visual Intelligence
In this report, we present PolyOCR, a family of unified OCR foundation models of varying scales.
ProofPaper ↗
Key points
- Optical Character Recognition (OCR) is evolving from plain-text transcription toward general visual intelligence, requiring models to recognize, localize, and reason over textual information in complex visual environments.
- PolyOCR combines a shared instruction-following framework with a large-scale data engine that converts heterogeneous visual resources into quality-verified OCR supervision.
- We introduce Competence-Guided Policy Optimization, which combines verifier-based Group Relative Policy Optimization with on-policy distillation through sample-wise routing based on teacher reliability and the teacher--student competence gap.
- We also introduce OCRBench v2.1, our revision of OCRBench v2 with manually verified annotation corrections and task-aligned scoring metrics.
Sources (1)
- [1]PolyOCR-Venus: Unified OCR Foundation Models for Text-Centric Visual IntelligenceHugging Face Daily Papers · Sep 29, 12:00 AM
In this report, we present PolyOCR, a family of unified OCR foundation models of varying scales.
Optical Character Recognition (OCR) is evolving from plain-text transcription toward general visual intelligence, requiring models to recognize, localize, and reason over textual information in complex visual environments.
Extractive summary: sentences quoted from the sources.
Before this
- Sep 28, 2026unslothai/unsloth v0.1.900-beta: Laya Decision Models + Library
- Sep 28, 2026Notes on NVIDIA Nemotron
- Aug 22, 2026sgl-project/sglang v0.5.18
- Aug 10, 2026huggingface/transformers v5.15.0: Release: v5.15.0