For example, one of the big wins was diamond (I think?) cutters for crystalline silicon, which allowed panels to be made rig far less inputs. A big competitor before that was thin film, and maybe you have heard of Solyndra, who had a heat tech that was made obsolete because this advancement in cutting silicon.
Any improvement along the chain of silicon production to siping to assembly has a chance to reduce costs, and it's usually small incremental bits of 1% here or 2% there.
The reason we are getting steady cost reduction is steady volume increase in production.
For decades we have been massively under-served in PV production volume. Even now, we are still somewhere in the middle of that S-curve.
I haven't seen such data, but panel in the early C20th probably followed a similar learning curve to PV now.