# VLCache: Computing 2% Vision Tokens and Reusing 98% for Vision-Language Inference

arXiv: https://arxiv.org/abs/2512.12977 · HTML: https://mlsanity.com/paper/2512.12977/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Builds on 17 of its 17 references that are in the map.
- First posted on arXiv 15 Dec 2025.

## Builds on (most cited first, 17 of 17)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [Qwen2.5-VL Technical Report](https://mlsanity.com/paper/2502.13923/) (2025)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [Qwen Technical Report](https://mlsanity.com/paper/2309.16609/) (2023)
- [Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond](https://mlsanity.com/paper/2308.12966/) (2023)
- [Qwen2 Technical Report](https://mlsanity.com/paper/2407.10671/) (2024)
- [SGLang: Efficient Execution of Structured Language Model Programs](https://mlsanity.com/paper/2312.07104/) (2023)
- [MMMU-Pro: A More Robust Multi-discipline Multimodal Understanding Benchmark](https://mlsanity.com/paper/2409.02813/) (2024)
- [Patch n' Pack: NaViT, a Vision Transformer for any Aspect Ratio and Resolution](https://mlsanity.com/paper/2307.06304/) (2023)
- [CacheBlend: Fast Large Language Model Serving for RAG with Cached Knowledge Fusion](https://mlsanity.com/paper/2405.16444/) (2024)
- [EPIC: Efficient Position-Independent Caching for Serving Large Language Models](https://mlsanity.com/paper/2410.15332/) (2024)
- [KVShare: An LLM Service System with Efficient and Effective Multi-Tenant KV Cache Reuse](https://mlsanity.com/paper/2503.16525/) (2025)
- [MPIC: Position-Independent Multimodal Context Caching System for Efficient MLLM Serving](https://mlsanity.com/paper/2502.01960/) (2025)

## Built on by (most cited first, 7 of 7)

- [CodecSight: Leveraging Video Codec Signals for Efficient Streaming VLM Inference](https://mlsanity.com/paper/2604.06036/) (2026)
- [Kamera: Unified Position-Invariant Multimodal KV Cache for Training-Free Reuse](https://mlsanity.com/paper/2606.23581/) (2026)
- [The Gate, Not the Cache: Gate Provenance Bounds the Closed-Loop Reliability of Training-Free VLA Token Skipping](https://mlsanity.com/paper/2608.00391/) (2026)
- [VLMaxxing through FrameMogging Training-Free Anti-Recomputation for Video Vision-Language Models](https://mlsanity.com/paper/2605.03351/) (2026)
- [Inverting the Streaming-Diffusion Bottleneck: Video-Rate MLLM-Conditioned Edit Diffusion on a Consumer GPU](https://mlsanity.com/paper/2606.05981/) (2026)
- [QV-PIC: Query-Aware Visual Position-Independent Caching for Efficient RAG Serving](https://mlsanity.com/paper/2608.12121/) (2026)
- [Visual Orchestration Tax in Agentic VLM Pipelines: Auditing and Certifying Visual Evidence Reuse](https://mlsanity.com/paper/2610.08170/) (2026)

## Abstract (arXiv)

This paper presents VLCache, a cache reuse framework that exploits both Key-Value (KV) cache and encoder cache from prior multimodal inputs to eliminate costly recomputation when the same multimodal inputs recur. Unlike previous heuristic approaches, we formally identify the cumulative reuse error effect and demonstrate how to minimize the non-prefix cache reuse error effectively. We further analyze the varying importance of model layers and propose a dynamic, layer-aware recomputation strategy to balance accuracy and efficiency. Experimental results show that VLCache achieves an accuracy on par with full recomputation, while requiring only 2-5% of the tokens to compute, yielding 1.2x-16x TTFT speedups. We develop an experimental implementation of the proposed VLCache pipeline based on SGLang, enabling significantly faster inference in practical deployments.