# CIVIC: End-to-End Sequence Compactness for Efficient Vision-Language Models

arXiv: https://arxiv.org/abs/2605.28115 · HTML: https://mlsanity.com/paper/2605.28115/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Visual token pruning and efficiency, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 18 of its 20 references that are in the map.
- First posted on arXiv 27 May 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Visual token pruning and efficiency](https://mlsanity.com/mllm/visual-token-pruning-and-efficiency/)
- Efficient LLM inference and serving / [Visual token pruning in multimodal models](https://mlsanity.com/inference/visual-token-pruning-in-multimodal-models/)

## Builds on (most cited first, 18 of 18)

- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification](https://mlsanity.com/paper/2106.02034/) (2021)
- [VisionZip: Longer is Better but Not Necessary in Vision Language Models](https://mlsanity.com/paper/2412.04467/) (2024)
- [TokenPacker: Efficient Visual Projector for Multimodal LLM](https://mlsanity.com/paper/2407.02392/) (2024)
- [LLaVA-Mini: Efficient Image and Video Large Multimodal Models with One Vision Token](https://mlsanity.com/paper/2501.03895/) (2025)
- [DiffRate : Differentiable Compression Rate for Efficient Vision Transformers](https://mlsanity.com/paper/2305.17997/) (2023)
- [VL-Cache: Sparsity and Modality-Aware KV Cache Compression for Vision-Language Model Inference Acceleration](https://mlsanity.com/paper/2410.23317/) (2024)
- [VFlowOpt: A Token Pruning Framework for LMMs with Visual Information Flow-Guided Optimization](https://mlsanity.com/paper/2508.05211/) (2025)
- [Learning to Merge Tokens via Decoupled Embedding for Efficient Vision Transformers](https://mlsanity.com/paper/2412.10569/) (2024)
- [InternVL-X: Advancing and Accelerating InternVL Series with Efficient Visual Token Compression](https://mlsanity.com/paper/2503.21307/) (2025)
- [VisionTrim: Unified Vision Token Compression for Training-Free MLLM Acceleration](https://mlsanity.com/paper/2601.22674/) (2026)
- [SparseVILA: Decoupling Visual Sparsity for Efficient VLM Inference](https://mlsanity.com/paper/2510.17777/) (2025)
- [PackCache: A Training-Free Acceleration Method for Unified Autoregressive Video Generation via Compact KV-Cache](https://mlsanity.com/paper/2601.04359/) (2026)
- [LightVLM: Acceleraing Large Multimodal Models with Pyramid Token Merging and KV Cache Compression](https://mlsanity.com/paper/2509.00419/) (2025)
- [LEO-MINI: An Efficient Multimodal Large Language Model using Conditional Token Reduction and Mixture of Multi-Modal Experts](https://mlsanity.com/paper/2504.04653/) (2025)
- [ZOO-Prune: Training-Free Token Pruning via Zeroth-Order Gradient Estimation in Vision-Language Models](https://mlsanity.com/paper/2509.24837/) (2025)
- [AirCache: Activating Inter-modal Relevancy KV Cache Compression for Efficient Large Vision-Language Model Inference](https://mlsanity.com/paper/2503.23956/) (2025)
- [HybridKV: Hybrid KV Cache Compression for Efficient Multimodal Large Language Model Inference](https://mlsanity.com/paper/2604.05887/) (2026)

## Abstract (arXiv)

Vision-Language Models (VLMs) face severe memory and latency bottlenecks due to high-resolution visual tokens. While current token reduction methods theoretically save FLOPs, post-hoc pruning introduces structural overhead, failing to yield proportional wall-clock acceleration. However, enforcing a contiguous compact pathway risks geometric disorientation and loss of fine-grained localization. To overcome these barriers, this paper introduces CIVIC, a path-consistent compact visual inference framework. By maintaining compact sequence representations seamlessly across the vision encoder, projection layer, LLM prefill, and KV-cache, CIVIC avoids non-contiguous memory access and localized unmerging overheads. Evaluated on the Qwen3-VL architecture, CIVIC successfully translates sequence reductions into genuine physical hardware efficiency, shrinking KV-cache memory to approximately one-third of the baseline and reducing end-to-end inference latency. Enabled by text-aligned KL distillation and an adaptive spatial retention floor, CIVIC achieves these efficiency milestones without degrading accuracy across rigorous multimodal reasoning and visual grounding benchmarks.