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

State-of-the-art sequence-to-sequence models often require autoregressivedecoding, which can be highly expensive. However, for some downstream taskssuch as out-of-distribution (OOD) detection and resource allocation, the actualdecoding output is not needed just a scalar attribute of this sequence. Inthese scenarios, where for example knowing the quality of a system's output topredict poor performance prevails over knowing the output itself, is itpossible to bypass the autoregressive decoding? We propose Non-AutoregressiveProxy (NAP) models that can efficiently predict general scalar-valuedsequence-level attributes. Importantly, NAPs predict these metrics directlyfrom the encodings, avoiding the expensive autoregressive decoding stage. Weconsider two sequence-to-sequence task: Machine Translation (MT); and AutomaticSpeech Recognition (ASR). In OOD for MT, NAPs outperform a deep ensemble whilebeing significantly faster. NAPs are also shown to be able to predictperformance metrics such as BERTScore (MT) or word error rate (ASR). Fordownstream tasks, such as data filtering and resource optimization, NAPsgenerate performance predictions that outperform predictive uncertainty whilebeing highly inference efficient.

Who Needs Decoders? Efficient Estimation of Sequence-level Attributes