AI Paper Atlas

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

Implicit Counterfactual Learning for Audio-Visual Segmentation

Mingfeng Zha, Tianyu Li, Guoqing Wang, Peng Wang, Yangyang Wu, Yang Yang, Heng Tao Shen

ICCV 2025 · Vision-language and multimodal › Cross-modal learning · approach: Audio-visual source localization · 9 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 all three components (MIT, SC, CDCL) to the baseline (ablation)ICF · AVS-Object M3J&F: 64.51 -> 69.89Improved
add all three components (MIT, SC, CDCL) to the baseline (ablation)ICF · AVSSJ&F: 41.86 -> 48.16Improved
add semantic counterfactual (SC) on top of MIT and CDCL (ablation)ICF · AVS-Object M3J&F: 67.63 -> 69.89Improved
add CDCL on top of MIT and SC (ablation)ICF · AVSSJ&F: 45.66 -> 48.16Improved
add video and segment granularity to frame-level implicit text (ablation)ICF · AVSSJ&F: 42.34 -> 44.10Improved
generate counterfactuals with diffusion in continuous space instead of discrete VQVAE (ablation)ICF · AVS-Object M3J&F: 68.51 -> 69.89Improved
add L_ortho constraint to continuous counterfactual generation (ablation)ICF · AVSSJ&F: 47.31 -> 48.16Improved
use inter-sample and intra-sample counterfactuals together instead of feature-level (ablation)ICF · AVS-Object M3J&F: 67.63 -> 69.89Improved
use distribution-based contrast instead of prototype contrast (ablation)ICF · AVSSJ&F: 46.22 -> 48.16Improved
replace implicit orthogonal counterfactuals with explicit Text Revision (ablation)ICF · AVS-Object M3J&F: 69.89 -> 68.44Hurt
integrate ICF into AVSBenchAVSBench · AVS-Object M3J&F: 59.25 -> 62.30Improved
integrate ICF into AVSegFormerAVSegFormer · AVSSJ&F: 39.33 -> 43.44Improved

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