values” or “goals

Can we predict model behavior in novel situations outside of the distribution of inputs we have access to with mechanistic understanding? a

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it

browse 1 citing papers

representative citing papers

cs.LG · 2025-01-27 · unverdicted · novelty 3.0

A review paper that organizes conceptual, practical, and socio-technical open problems in mechanistic interpretability.

Showing 1 of 1 citing paper.

Open Problems in Mechanistic Interpretability cs.LG · 2025-01-27 · unverdicted · none · ref 49
A review paper that organizes conceptual, practical, and socio-technical open problems in mechanistic interpretability.