mlsanityData to 7 Oct 2026

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

arXiv 2605.28115 ↗First posted 27 May 2026cs.AIcited by 0

On the map: Multimodal LLMs / vision-language models → Visual token pruning and efficiency

also in 1 more
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.

Key facts. No citing papers in the map yet (data through 7 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.

Builds on

18 of 20 references.

Closest ancestors · 13

Same direction, newest first

  1. 2026HybridKV: Hybrid KV Cache Compression for Efficient Multimodal Large Language Model Inference7 Apr
  2. 2026VisionTrim: Unified Vision Token Compression for Training-Free MLLM Acceleration30 Jan
  3. 2025SparseVILA: Decoupling Visual Sparsity for Efficient VLM Inference20 Oct
  4. 2025ZOO-Prune: Training-Free Token Pruning via Zeroth-Order Gradient Estimation in Vision-Language Models29 Sep
  5. 2025LightVLM: Acceleraing Large Multimodal Models with Pyramid Token Merging and KV Cache Compression30 Aug
  6. 2025VFlowOpt: A Token Pruning Framework for LMMs with Visual Information Flow-Guided Optimization7 Aug
  7. 2025LEO-MINI: An Efficient Multimodal Large Language Model using Conditional Token Reduction and Mixture of Multi-Modal Experts7 Apr
  8. 2025AirCache: Activating Inter-modal Relevancy KV Cache Compression for Efficient Large Vision-Language Model Inference31 Mar
  9. 2025InternVL-X: Advancing and Accelerating InternVL Series with Efficient Visual Token Compression27 Mar
  10. 2025LLaVA-Mini: Efficient Image and Video Large Multimodal Models with One Vision Token7 Jan
  11. 2024VisionZip: Longer is Better but Not Necessary in Vision Language Models5 Dec
  12. 2024VL-Cache: Sparsity and Modality-Aware KV Cache Compression for Vision-Language Model Inference Acceleration29 Oct
1 more
  1. 2024TokenPacker: Efficient Visual Projector for Multimodal LLM2 Jul

Other references · 5

  1. 2022Token Merging: Your ViT But Fastercited by 995
  2. 2021DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsificationcited by 849
  3. 2023DiffRate : Differentiable Compression Rate for Efficient Vision Transformerscited by 78
  4. 2024Learning to Merge Tokens via Decoupled Embedding for Efficient Vision Transformerscited by 21
  5. 2026PackCache: A Training-Free Acceleration Method for Unified Autoregressive Video Generation via Compact KV-Cachecited by 9

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Efficient Inference for Large Vision-Language Models: Bottlenecks, Techniques, and Prospects
  2. 2026Empirical Recipes for Efficient and Compact Vision-Language Models
  3. 2025KV-Efficient VLA: A Method to Speed up Vision Language Models with RNN-Gated Chunked KV Cache
  4. 2026Towards Efficient Large Vision-Language Models: A Comprehensive Survey on Inference Strategies
  5. 2025VScan: Rethinking Visual Token Reduction for Efficient Large Vision-Language Models
  6. 2024Inference Optimal VLMs Need Fewer Visual Tokens and More Parameters
  7. 2025Efficient Vision-Language Reasoning via Adaptive Token Pruning
  8. 2026PACE: A Unified Condense-and-Extract Paradigm for Fast VLM Inference