The gradient flow for linear in-context learning is derived in closed form, and two special cases are classified into attracting minima, saddle points, and invariant manifolds.
Trans- formers learn to implement preconditioned gradient descent for in-context learning
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
math.DS 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Dynamical Behaviors of the Gradient Flows for In-Context Learning
The gradient flow for linear in-context learning is derived in closed form, and two special cases are classified into attracting minima, saddle points, and invariant manifolds.