Impact regularizer (defined) — borrows its improvement guarantee from → Robust policy improvement

Reachability analysis tells the impact regularizer which states are safe to fall back to; robust policy improvement tells it how to leave that fallback without losing the safety guarantee. The paper pairs the two in one sentence when describing how to replace the null baseline with "a known safe... but suboptimal policy, and then seek to improve the policy from there, somewhat reminiscent of reachability analysis... or robust policy improvement" (concrete-problems, §"Define an Impact Regularizer:", p. 5). The distinction matters for what the regularizer can do: reachability analysis defines the boundary of acceptable states, while robust policy improvement supplies the guarantee that optimizing away from the passive baseline will not perform worse than that baseline even under model uncertainty. One names the region; the other licenses movement within it.