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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

To build an artificial neural network like the biological intelligencesystem, recent works have unified numerous tasks into a generalist model, whichcan process various tasks with shared parameters and do not have anytask-specific modules. While generalist models achieve promising results onvarious benchmarks, they have performance degradation on some tasks comparedwith task-specialized models. In this work, we find that interference amongdifferent tasks and modalities is the main factor to this phenomenon. Tomitigate such interference, we introduce the Conditional Mixture-of-Experts(Conditional MoEs) to generalist models. Routing strategies under differentlevels of conditions are proposed to take both the training/inference cost andgeneralization ability into account. By incorporating the proposed ConditionalMoEs, the recently proposed generalist model Uni-Perceiver can effectivelymitigate the interference across tasks and modalities, and achievesstate-of-the-art results on a series of downstream tasks via prompt tuning on1% of downstream data. Moreover, the introduction of Conditional MoEs stillholds the generalization ability of generalist models to conduct zero-shotinference on new tasks, e.g., video-text retrieval and video caption. Code andpre-trained generalist models shall be released.

Uni-Perceiver-MoE: Learning Sparse Generalist Models with Conditional MoEs