English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Multi-Task Learning (MTL) is designed to train multiple correlated taskssimultaneously, thereby enhancing the performance of individual tasks.Typically, a multi-task network structure consists of a shared backbone andtask-specific decoders. However, the complexity of the decoders increases withthe number of tasks. To tackle this challenge, we integrate the decoder-freevision-language model CLIP, which exhibits robust zero-shot generalizationcapability. Recently, parameter-efficient transfer learning methods have beenextensively explored with CLIP for adapting to downstream tasks, where prompttuning showcases strong potential. Nevertheless, these methods solely fine-tunea single modality (text or visual), disrupting the modality structure of CLIP.In this paper, we first propose Multi-modal Alignment Prompt (MmAP) for CLIP,which aligns text and visual modalities during fine-tuning process. Buildingupon MmAP, we develop an innovative multi-task prompt learning framework. Onthe one hand, to maximize the complementarity of tasks with high similarity, weutilize a gradient-driven task grouping method that partitions tasks intoseveral disjoint groups and assign a group-shared MmAP to each group. On theother hand, to preserve the unique characteristics of each task, we assign antask-specific MmAP to each task. Comprehensive experiments on two largemulti-task learning datasets demonstrate that our method achieves significantperformance improvements compared to full fine-tuning while only utilizingapproximately 0.09% of trainable parameters.

MmAP : Multi-modal Alignment Prompt for Cross-domain Multi-task Learning