OneVision-Encoder: Codec-Aligned Sparsity as a Foundational Principle for Multimodal Intelligence

2026-02-09

AI Summary

This work addresses the inefficiency of current vision models, which uniformly process redundant pixel grids and struggle to focus on semantically critical regions. Inspired by video codecs, the authors propose a sparse computation paradigm guided by codec-aligned sparsity principles, activating computation only in high-information-entropy regions (3.1%–25% of the input). The approach integrates Codec Patchification, shared 3D RoPE positional encoding, and discriminative training with million-scale semantic clustering to unify spatiotemporal modeling. Evaluated across 16 benchmarks spanning image, video, and document understanding, the method outperforms strong baselines such as Qwen3-ViT and SigLIP2, achieving an average 4.1% gain on video tasks while using fewer visual tokens and less pretraining data—demonstrating a positive correlation between efficiency and accuracy.

Technology Category

• Computer Vision: Large Vision Models
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Hypothesis. Artificial general intelligence is, at its core, a compression problem. Effective compression demands resonance: deep learning scales best when its architecture aligns with the fundamental structure of the data. These are the fundamental principles. Yet, modern vision architectures have strayed from these truths: visual signals are highly redundant, while discriminative information, the surprise, is sparse. Current models process dense pixel grids uniformly, wasting vast compute on static background rather than focusing on the predictive residuals that define motion and meaning. We argue that to solve visual understanding, we must align our architectures with the information-theoretic principles of video, i.e., Codecs. Method. OneVision-Encoder encodes video by compressing predictive visual structure into semantic meaning. By adopting Codec Patchification, OV-Encoder abandons uniform computation to focus exclusively on the 3.1%-25% of regions rich in signal entropy. To unify spatial and temporal reasoning under irregular token layouts, OneVision-Encoder employs a shared 3D RoPE and is trained with a large-scale cluster discrimination objective over more than one million semantic concepts, jointly capturing object permanence and motion dynamics. Evidence. The results validate our core hypothesis: efficiency and accuracy are not a trade-off; they are positively correlated. When integrated into LLM, it consistently outperforms strong vision backbones such as Qwen3-ViT and SigLIP2 across 16 image, video, and document understanding benchmarks, despite using substantially fewer visual tokens and pretraining data. Notably, on video understanding tasks, OV-Encoder achieves an average improvement of 4.1% over Qwen3-ViT. Codec-aligned, patch-level sparsity is a foundational principle, enabling OV-Encoder as a scalable engine for next-generation visual generalists.

Problem

• sparsity
• video compression
• visual redundancy
• predictive residuals
• codec alignment

Innovation

• Codec-Aligned Sparsity
• Predictive Residual Focus
• 3D RoPE
• Irregular Token Layout
• Cluster Discrimination