AI Paper Atlas

Research map / Video / Action recognition

${\text{CA}^{2}\text{ST}}$: Cross-Attention in Audio, Space, and Time for Holistic Video Recognition

Jongseo Lee, Joohyun Chang, Dongho Lee, Jinwoo Choi

TPAMI 2025 · Video › Action recognition · approach: Contrastive video representation learning · 6 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 B-CA information exchange between experts (ablation)independent experts w/ adapter · EK100 actiontop-1 accuracy: 41.7 -> 48.7Improved
add bottleneck adapters in B-CA (ablation)CAST · EK100 actiontop-1 accuracy: 39.4 -> 48.7Improved
place adapters before cross-attention (B-CA) instead of after (ablation)CAST · EK100 actiontop-1 accuracy: 47.9 -> 48.7Improved
use bi-directional cross-attention instead of S2T only (ablation)CAST · EK100 actiontop-1 accuracy: 43.7 -> 48.7Improved
use bi-directional cross-attention instead of T2S only (ablation)CAST · EK100 actiontop-1 accuracy: 46.7 -> 48.7Improved
use cross-attention instead of lateral connection (ablation)CAST · EK100 actiontop-1 accuracy: 39.0 -> 48.7Improved
time-to-space window shape instead of space-time for both (ablation)CAST · EK100 actiontop-1 accuracy: 47.2 -> 48.7Improved
use CLIP spatial and VideoMAE temporal experts instead of VideoMAE for both (ablation)CAST · EK100 actiontop-1 accuracy: 40.3 -> 48.7Improved
swap roles of CLIP and VideoMAE experts (ablation)CAST · EK100 actiontop-1 accuracy: 48.7 -> 47.8No clear effect
add cross-attention exchange between audio and visual experts (ablation)CAVA · EPIC-SOUNDStop-1 accuracy: 55.6 -> 60.3Improved
use B-CA instead of late fusion with adapter (ablation)CAVA · EPIC-SOUNDStop-1 accuracy: 57.9 -> 60.3Improved
layer-wise connection without cross-attention instead of late fusion (ablation)CAVA · EPIC-SOUNDStop-1 accuracy: 57.9 -> 55.2Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 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