Emergence of Abstractions: Concept Encoding and Decoding Mechanism for In-Context Learning in Transformers

Bibliografske podrobnosti

izdano v:	arXiv.org (Dec 18, 2024), p. n/a
Glavni avtor:	Han, Seungwook
Drugi avtorji:	Song, Jinyeop, Gore, Jeff, Agrawal, Pulkit
Izdano:	Cornell University Library, arXiv.org
Teme:	Predictive control Algorithms Encoding-Decoding Large language models Context Failure modes Task complexity Representations Adaptive learning
Online dostop:	Citation/Abstract Full text outside of ProQuest
Oznake:	Označite Brez oznak, prvi označite!