In (effective) field theories, the lagrangian has a field
redefinition redundancy, meaning many different lagrangians can generate
the same amplitudes. Just like gauge redundancy in a gauge theory, this
makes the map from lagrangians to amplitudes difficult: observable
physics is obscured behind a combinatorially large set of Feynman
diagrams, each of them unphysical, and of which pieces will generally
cancel against each other.
I will describe briefly how - in a scalar field theory - concepts from
geometry can cut through this noise by a) identifying field redefinition
invariant quantities in lagrangians that map neatly onto coefficients in
amplitudes and b) identifying one of the many equivalent lagrangians
that is objectively better for calculating amplitudes.