AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Efficient Multimodal Large Language Model via Dynamic KV Cache Quantization

Jiahao Fan, Chien-Ming Chen

AAAI 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Training-free token reduction for MLLMs · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a dynamic KV cache quantization approach to reduce the memory and compute cost of multimodal LLM inference for resource-constrained deployment.

Problem

Multimodal LLMs face high computational cost and KV-cache memory footprint during inference, hindering real-time deployment on resource-constrained devices.

Approach

Applies dynamic quantization to the KV cache of multimodal LLMs to shrink its memory footprint during inference; specific mechanism not stated in the abstract.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add fine-grained grouping (ablation)Dynamic KV Cache Quantization (LLaVA-1.5-7B) · TextVQAaccuracy (%): 57.64 -> 58.01Improved
add fine-grained grouping (ablation)Dynamic KV Cache Quantization (LLaVA-1.5-7B) · GQAaccuracy (%): 61.47 -> 62.13Improved
add token/channel recorders and scaling (ablation)Dynamic KV Cache Quantization (LLaVA-1.5-7B) · GQAaccuracy (%): 61.75 -> 62.13Improved
add token/channel recorders and scaling (ablation)Dynamic KV Cache Quantization (LLaVA-1.5-7B) · TextVQAaccuracy (%): 57.82 -> 58.01Improved
decrease group size from 128 to 32 (ablation)Dynamic KV Cache Quantization (LLaVA-1.5-7B) · TextVQAaccuracy (%): 57.43 -> 58.01Improved
decrease group size from 32 to 16 (ablation)Dynamic KV Cache Quantization (LLaVA-1.5-7B) · TextVQAaccuracy (%): 58.01 -> 58.05No clear effect
use per-token instead of per-channel grouping for K (ablation)Dynamic KV Cache Quantization (LLaVA-1.5-7B) · GQAaccuracy (%): 61.92 -> 62.05Mixed
use per-token instead of per-channel grouping for V (ablation)Dynamic KV Cache Quantization (LLaVA-1.5-7B) · GQAaccuracy (%): 61.84 -> 62.13Improved
use Dynamic KV Cache Quantization instead of uniform 4-bit KV quantizationDynamic KV Cache Quantization (LLaVA-1.5-7B) · TextVQA (LLaVA-1.5-7B)accuracy (%): 0.12 -> 58.01Improved
use Dynamic KV Cache Quantization instead of SKVQDynamic KV Cache Quantization (LLaVA-1.5-7B) · SciQA (LLaVA-1.5-7B)accuracy (%): 56.02 -> 69.74Improved
use Dynamic KV Cache Quantization instead of KIVIDynamic KV Cache Quantization (LLaVA-1.5-7B) · TextVQA (Qwen-VL)accuracy (%): 63.87 -> 63.92No clear effect
use Dynamic KV Cache Quantization instead of MiKVDynamic KV Cache Quantization (LLaVA-1.5-7B) · TextVQA (LLaVA-1.5-13B)accuracy (%): 60.87 -> 61.12No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy