DeepSeek-OCR 2: Visual Causal Flow

TL;DR

DeepSeek-OCR 2 introduces a novel encoder capable of dynamically reordering visual tokens based on semantic and causal reasoning, aiming to improve 2D image understanding by mimicking human visual perception.

Contribution

It proposes a new encoder architecture with causal reasoning to reorder visual tokens, enhancing 2D image understanding in vision-language models.

Findings

Enables causal-informed token reordering for complex layouts

Demonstrates improved semantic coherence in visual token processing

Provides publicly accessible code and models

Abstract

We present DeepSeek-OCR 2 to investigate the feasibility of a novel encoder-DeepEncoder V2-capable of dynamically reordering visual tokens upon image semantics. Conventional vision-language models (VLMs) invariably process visual tokens in a rigid raster-scan order (top-left to bottom-right) with fixed positional encoding when fed into LLMs. However, this contradicts human visual perception, which follows flexible yet semantically coherent scanning patterns driven by inherent logical structures. Particularly for images with complex layouts, human vision exhibits causally-informed sequential processing. Inspired by this cognitive mechanism, DeepEncoder V2 is designed to endow the encoder with causal reasoning capabilities, enabling it to intelligently reorder visual tokens prior to LLM-based content interpretation. This work explores a novel paradigm: whether 2D image understanding can be effectively achieved through two-cascaded 1D causal reasoning structures, thereby offering a new architectural approach with the potential to achieve genuine 2D reasoning. Codes and model weights are publicly accessible at http://github.com/deepseek-ai/DeepSeek-OCR-2.