Conference Open access

MLDA: Test-Time Multi-Level Adaptation with Dynamic Alignment for Compositional Zero-Shot Learning

Miao-Ge Li Yu Liu Jingcai Guo

Sep 2026 · Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence · 0 citations · 42 references

TL;DR

MLDA is introduced, a test-time adaptation framework for CZSL that jointly derives multi-level prototypes and performs dynamic vision–language alignment and leverages optimal transport to achieve consistent matching of semantically salient information across modalities.

Abstract

Compositional Zero-Shot Learning (CZSL) aims to recognize novel attribute–object compositions by leveraging shared knowledge from seen primitives. While recent works have begun to incorporate test-time adaptation to mitigate performance degradation caused by label-space distribution shifts, they suffer from primitive-level knowledge underutilization and cross-modal semantic asymmetry, hindering robust understanding of novel compositions. To address these issues, we introduce MLDA, a test-time adaptation framework for CZSL that jointly derives multi-level prototypes and performs dynamic vision–language alignment. Specifically, we construct and progressively update composition- and primitive-level prototype sets to accumulate structured semantic knowledge. By augmenting test inputs with diverse visual perspectives and enriched compositional descriptions, MLDA leverages optimal transport to achieve consistent matching of semantically salient information across modalities. Furthermore, composition activation scores are dynamically updated over the test stream to emphasize informative compositions and suppress irrelevant ones. Extensive experiments are conducted to verify the effectiveness of the proposed method. The code is available at https://github.com/keepgoingjkg/MLDA.

DOI Semantic Scholar

Read PDF

Similar papers

Preprint Aug 2026

DIFFCZSL: Compositional Zero-Shot Learning Regularized by Diffusion Representations

DIFFCZSL is proposed, a diffusion-augmented framework that injects generative priors from pre-trained diffusion models into CLIP-based CZSL pipelines and highlights the complementary strengths of generative diffusion representations and discriminative vision-language models for compositional generalization.

Hangyu Tian, Zhen-Qi He, Yanghao Wang et al. · 0 citations

Open access Jul 2026

Decoupled Semantic Flow Matching with Dynamic Reconstruction for Generalized Zero-Shot Learning

SemanticFlowNet is introduced, a framework based on Decoupled Semantic Flow Matching that relies on channel-wise concatenation of temporal encodings, semantic attributes, and visual contexts, which preserves the orthogonal subspaces of each modality and reduces interference.

Chu-Yang Song, Mingyi Song, Yang Liu et al. · Electronics · 0 citations

Preprint Aug 2026

MLLMCLIP: Feature-Level Distillation of MLLM for Robust Vision-Language Representations

Compared to prior CLIP-enhancement methods, MLLMCLIP achieves state-of-the-art compositional accuracy while delivering consistent gains on standard zero-shot classification and image-text retrieval, showing that feature-level distillation strengthens both compositional and general vision-language representation capabil...

Jongsuk Kim, Qi-Yu Wu, Zhuoyuan Mao et al. · 0 citations

#artificial intelligence Preprint Aug 2026

From Inference to Adaptation: A Unified Optimal Transport View of Vision Language Model

This work proposes a principled VLM TTA method called \algname, and theoretically reveals that the InfoNCE loss can be neatly reformulated as a Wasserstein OT formulation, thereby unifying the objectives of the inference and adaptation of VLMs to achieve their mutual benefits.

Qi Yu, Zhichen Zeng, Katherine Tieu et al. · 0 citations

Preprint Sep 2026

From Model Patterns to Abstract Semantics in Compositional Zero-Shot Learning

This work presents a new perspective that views primitive variations as the context-driven activation of concrete visual cues rather than independent entities rather than independent entities, and proposes CLEAR, a CLoze-style rEAsoning-based Re-ranking framework inspired by human perceptual processes.

Wei-Ze Li, Zhi-Cheng Zhao, Fei Su · 0 citations

Conference Open access Sep 2026

CoDA: Co-adaptive Dual-path Alignment for Vision-Language Models

In CoDA, a new adaptation framework that explicitly disentangles and coordinates cross-modal semantic alignment and intra-modal structural consistency is proposed, and it is shown that CoDA outperforms state-of-the-art parameter-efficient methods, particularly under few-shot learning and distribution-shift scenarios.

Yi Zhang, Rui Zhu, Chan-Ni Li et al. · Proceedings of the Thirty-Fi... · 0 citations