finding
active
finding:code-domain-shows-smooth-predictive-trajectory-5-4-versus-novel-saes-46-218

Code domain shows smooth predictive trajectory (τ=5.4) versus novel/SAEs (τ=46–218).

Fig. 26.

Source paper

extracted_from
Priors in Time: Missing Inductive Biases for Language Model Interpretability
(2025) · Lubana, Ekdeep Singh · Rager, Can · Hindupur, Sai Sumedh R. · Costa, Valerie +12

Neighborhood — ranked by edge-count