Clarify tensor-valued object treatment in README

Updated the README to clarify the treatment of inputs, outputs, sensitivities, and uncertainties.

Author: heptaflar

README.md

@@ -14,8 +14,9 @@ linearizations directly from the implementation whenever needed, so sensitivity
 information automatically stays consistent with the code. Combined with random
 sampling methods for strongly nonlinear behaviour, this defines uncertainty
 propagation in terms of algorithmically differentiable programs. AD frameworks
-treat inputs, outputs, and uncertainties as tensor‑valued objects rather than
-forcing the uncertainty calculus into a fixed set of closed‑form formulas.
+treat inputs, outputs, sensitivities, and uncertainties as dynamic tensor‑valued
+objects rather than forcing the uncertainty calculus into a fixed set of
+closed‑form formulas.
 
 The ideas presented here grew out of earlier project-specific implementations
 of AD-based uncertainty propagation for harmonised satellite calibration