AI Paper Atlas

Research map / Video / Video-language understanding

AccKV: Towards Efficient Audio-Video LLMs Inference via Adaptive-Focusing and Cross-Calibration KV Cache Optimization

Zhonghua Jiang, Kui Chen, Kunxi Li, Keting Yin, Yiyun Zhou, Zhaode Wang, Chengfei Lv, Shengyu Zhang

AAAI 2026 · Video › Video-language understanding · approach: Video LLMs with fine-grained understanding · 5 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add Cross-Calibration (intra-modal merge plus cross-modal alignment) (ablation)AccKV · AVSD (VideoLLaMA2, 20% budget)ROUGE: 0.364630 -> 0.425179Improved
add Adaptive-Focusing (ablation)AccKV · AVSD (VideoLLaMA2, 20% budget)ROUGE: 0.342471 -> 0.364630Improved
add Attention Redistribution (ablation)AccKV · AVSD (VideoLLaMA2, 20% budget)ROUGE: 0.334883 -> 0.342471Improved
evict video KV in high layers instead of audio KV (ablation)VideoLLaMA2 · Audio tasksaccuracy: 0.691272 -> 0.523281Hurt
merge unimportant audio and video KV across modalities instead of evicting (ablation)VideoLLaMA2 · Audio tasksaccuracy: 0.651984 -> 0.166571Hurt
evict all audio KV cache (ablation)VideoLLaMA2 · Audio tasksaccuracy: 0.700490 -> 0.018601Hurt
evict all video KV cache (ablation)VideoLLaMA2 · Video tasksaccuracy: 0.541667 -> 0.09015Hurt
lower cache budget from 20% to 10% under AccKV (tau=0.9)AccKV · AVSD (VideoLLaMA2)ROUGE: 0.425179 -> 0.390516Hurt
use AccKV instead of H2O at 10% budgetVideoLLaMA2 · AVSD (10% budget)ROUGE: 0.332905 -> 0.390516Improved
use AccKV instead of SnapKV at 20% budgetVideoLLaMA2 · AVSD (20% budget)ROUGE: 0.424881 -> 0.425179No clear effect
lower threshold tau from 0.9 to 0.6 (ablation)AccKV · AVSD (VideoLLaMA2, 20% budget)ROUGE: 0.425179 -> 0.423846No clear effect
use AccKV instead of FastV at 120 token budgetAVicuna · AVSD (120 tokens)ROUGE: 0.213519 -> 0.315302Improved

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