In-context learning emerges as implicit Bayesian inference of latent concepts when pretraining data has long-range coherence, proven for mixture-of-HMM distributions and replicated on the synthetic GINC dataset.
Adversarially robust generalization requires more data , year =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2021 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
An Explanation of In-context Learning as Implicit Bayesian Inference
In-context learning emerges as implicit Bayesian inference of latent concepts when pretraining data has long-range coherence, proven for mixture-of-HMM distributions and replicated on the synthetic GINC dataset.