AI Paper Atlas

Research map / Vision-language and multimodal / Cross-modal learning

CLASP: Cross-modal Salient Anchor-based Semantic Propagation for Weakly-supervised Dense Audio-Visual Event Localization

Jinxing Zhou, Ziheng Zhou, Yanghao Zhou, Yuxin Mao, Zhangling Duan, Dan Guo

AAAI 2026 · Vision-language and multimodal › Cross-modal learning · approach: Self-supervised audio-visual representation learning · 4 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 anchor features to temporal propagation (ATP) (ablation)CLASP · UnAV-100 (weakly supervised)Avg. mAP: 29.3 -> 30.0Improved
add Global Anchor Identification to Local Anchor Identification (ablation)CLASP · UnAV-100 (weakly supervised)Avg. mAP: 29.8 -> 30.0Improved
add Local Anchor Identification to Global Anchor Identification (ablation)CLASP · UnAV-100 (weakly supervised)Avg. mAP: 29.5 -> 30.0Improved
compute agreement score from raw backbone features instead of unimodal-transformer features (ablation)CLASP MEAE · UnAV-100 (weakly supervised)Avg. mAP: 29.6 -> 30.0Improved
increase global anchor number K from 5 to 10 (ablation)CLASP · UnAV-100 (weakly supervised)Avg. mAP: 29.4 -> 30.0Improved
increase global anchor number K from 10 to 20 (ablation)CLASP · UnAV-100 (weakly supervised)Avg. mAP: 30.0 -> 29.1Hurt
increase local anchor number k from 4 to 8 (ablation)CLASP · UnAV-100 (weakly supervised)Avg. mAP: 30.0 -> 29.3Hurt
increase local anchor number k from 1 to 4 (ablation)CLASP · UnAV-100 (weakly supervised)Avg. mAP: 29.2 -> 30.0Improved
change local window number M from 14 to 4 or 28 (ablation)CLASP · UnAV-100 (weakly supervised)Avg. mAP: 30.0 -> 29.4 (M=4), 29.6 (M=28)Hurt
use CLASP instead of fully-supervised CCNet trained weaklyW-DAVEL · UnAV-100 (weakly supervised)Avg. mAP: 26.9 -> 30.0Improved
use CLASP instead of CCNetW-DAVEL · ActivityNet1.3 (weakly supervised)Avg. mAP: 19.6 -> 21.3Improved
use CLASP instead of CCNet for efficiencyW-DAVEL · UnAV-100 (weakly supervised)Parameters (M): 238.8 -> 27.2Improved

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