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

Large language models (LLMs) exploit in-context learning (ICL) to solve taskswith only a few demonstrations, but its mechanisms are not yet well-understood.Some works suggest that LLMs only recall already learned concepts frompre-training, while others hint that ICL performs implicit learning overdemonstrations. We characterize two ways through which ICL leveragesdemonstrations. Task recognition (TR) captures the extent to which LLMs canrecognize a task through demonstrations -- even without ground-truth labels --and apply their pre-trained priors, whereas task learning (TL) is the abilityto capture new input-label mappings unseen in pre-training. Using a wide rangeof classification datasets and three LLM families (GPT-3, LLaMA and OPT), wedesign controlled experiments to disentangle the roles of TR and TL in ICL. Weshow that (1) models can achieve non-trivial performance with only TR, and TRdoes not further improve with larger models or more demonstrations; (2) LLMsacquire TL as the model scales, and TL's performance consistently improves withmore demonstrations in context. Our findings unravel two different forcesbehind ICL and we advocate for discriminating them in future ICL research dueto their distinct nature.

What In-Context Learning "Learns" In-Context: Disentangling Task Recognition and Task Learning