← Search

Grégoire Delétang

2 accepted papers

2023

Randomized Positional Encodings Boost Length Generalization of Transformers

ACL 2023short

Transformers have impressive generalization capabilities on tasks with a fixed context length. However, they fail to generalize to sequences of arbitrary length, even for seemingly simple tasks such as duplicating a string. Moreover, simply training on longer sequences is inefficient due to the quad…

2020

Meta-trained agents implement Bayes-optimal agents

NeurIPS 2020spotlight

Memory-based meta-learning is a powerful technique to build agents that adapt fast to any task within a target distribution. A previous theoretical study has argued that this remarkable performance is because the meta-training protocol incentivises agents to behave Bayes-optimally. We empirically in…

Cited by 49SourcePDFScholar