AI Paper Atlas

Research map / Vision-language and multimodal / CLIP and zero-shot

Causality-guided Prompt Learning for Vision-language Models via Visual Granulation

Mengyu Gao, Qiulei Dong

ICCV 2025 · Vision-language and multimodal › CLIP and zero-shot · approach: Prompt learning for CLIP · 3 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 BBDM-based attribute disentanglement (ablation)CaPL · Base-to-new ImageNet-1Kharmonic mean H: 75.87 -> 77.80Improved
add factual intervention (ablation)CaPL · Base-to-new ImageNet-1Kharmonic mean H: 75.40 -> 77.80Improved
add counterfactual intervention (ablation)CaPL · Base-to-new ImageNet-1Kharmonic mean H: 75.42 -> 77.80Improved
add non-individualized attributes to factual granules (ablation)CaPL · Base-to-new ImageNet-1Kharmonic mean H: 76.53 -> 77.80Improved
use BBDM instead of classification-based disentanglement (ablation)CaPL · Base-to-new ImageNet-1Kharmonic mean H: 76.28 -> 77.80Improved
use BBDM instead of DDPM for disentanglement (ablation)CaPL · Base-to-new ImageNet-1Kharmonic mean H: 76.97 -> 77.80Improved
use non-individualized as source and individualized as condition (vs swapped) (ablation)CaPL · Base-to-new ImageNet-1Kharmonic mean H: 77.05 -> 77.80Improved
learnable random-initialized attribute queries instead of fixed GPT-3 queries (ablation)CaPL · Base-to-new ImageNet-1Kharmonic mean H: 75.20 -> 77.80Improved
random instead of GPT-3 prior initialization of queries (ablation)CaPL · Base-to-new ImageNet-1Kharmonic mean H: 76.17 -> 77.80Improved
increase number of individualized attributes K from 0 to 10 (ablation)CaPL · Base-to-new ImageNet-1Kharmonic mean H: 75.66 -> 77.80Improved
increase K from 10 to 12 (ablation)CaPL · Base-to-new ImageNet-1Kharmonic mean H: 77.80 -> 77.64No clear effect
add factual loss weight lambda_f=1 (vs 0) (ablation)CaPL · Base-to-new ImageNet-1Kharmonic mean H: 76.99 -> 77.80Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 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