Decompose and Conquer: Compositional Reasoning for Zero-Shot Temporal Action Localization
Haoyu Tang, Tianyuan Liang, Han Jiang, Xuesong Liu, Qinghai Zheng, Yupeng Hu
Abstract
Current Zero-Shot Temporal Action Localization (ZSTAL) methods, whether training-based or training-free ones, still predominantly rely on a single, unified query to localize an entire action. This unified representation is fundamentally ill-suited for complex real-world activities, as it fails to capture their internal compositional structure and adapt to dynamic, multi-stage variations across videos. To address this, we regard ZSTAL as a compositional reasoning task and introduce CASCADE, a Context-Aware Staged Action DEcomposition framework. Inspired by the human cognitive process of perceiving context, decomposing events, and reconstructing instances, CASCADE follows a training-free pipeline. It first perceives the video
BibTeX
@inproceedings{aaai2026_decomposeandconq,
title = {Decompose and Conquer: Compositional Reasoning for Zero-Shot Temporal Action Localization},
author = {Haoyu Tang and Tianyuan Liang and Han Jiang and Xuesong Liu and Qinghai Zheng and Yupeng Hu},
booktitle = {AAAI 2026},
year = {2026}
}