I understand that std::move unconditionally casts an lvalue to an rvalue. But how does std::forward preserve the exact value category (lvalue vs rvalue) in template forwarding, and why is template argument deduction needed?

std::forward<T>(arg) relies on Reference Collapsing rules! When T is an lvalue (e.g. Widget&), T& && collapses to Widget&, preserving the lvalue. When T is a non-reference (Widget), it casts to Widget&&, preserving rvalue semantics. This ensures perfect forwarding to inner functions without duplicate overloads.

A simple rule of thumb: use std::move on rvalue references (T&& arg with concrete type T), and use std::forward on forwarding references (T&& arg where T is a deduced template parameter)!