Announcing the ARC White-Box Estimation Challenge — Alignment Research Center

Research question

Can internal access improve estimates of neural-network expectations?

Design & assumptions

A compute-constrained challenge estimates random MLP expected outputs using mean-squared error.

Findings

White-box methods are motivated by limits of behavioral sampling.

Current methods degrade with depth, and trained-network transfer remains unresolved.

Limits & counterevidence

Initial-announcement configuration only; later contest rounds may differ and current round is unverified. Foundational estimation challenge, not a frontier-model safety score.

Source evidence