I've always tried to enthusiastically "sit through" such sessions but unless the field is really really close to my own, I don't understand anything.
Like their whole literature or notations feels alien language to me.
In one of such cases, I wanted to know about "differentiable programming languages" which was posted here recently. I watched this seminar (https://www.youtube.com/watch?v=x1-YOWa0vJU) but I didn't understand anything and lost all my confidence in what I knew.
(Note that I searched the presenter, he is very much respected in his field. I just provided that clip as an example. It's not just this, as I said, I feel the same in 90% of similar topics)
I remembered an MIT colloquium that took place in the late fifties; it was one of the first I attended at MIT. The speaker was Eugenio Calabi. Sitting in the front row of the audience were Norbert Wiener, asleep as usual until the time came to applaud, and Dirk Struik, who had been one of Calabi’s teachers when Calabi was an under-graduate at MIT in the forties. The subject of the lecture was beyond my competence. After the first five minutes I was completely lost. At the end of the lecture an arcane dialogue took place between the speaker and some members of the audience—Ambrose and Singer if I remember correctly. There followed a period of tense silence. Professor Struik broke the ice. He raised his hand and said, “Give us something to take home!” Calabi obliged, and in the next five minutes he explained in beautiful simple terms the gist of his lecture. Everybody filed out with a feeling of satisfaction.
They'd be just as lost in a presentation by a PhD in your field explaining their work.
That's before you get to the fact that most such presentations aren't very good. Presentations are yet another skill unto themselves, and few PhDs have actually mastered it. I think of them as more like infomercials. If you actually want to get on board with the material, you'll go read the paper or talk with the people. Or better, both.
There's a famous cartoon putting this into perspective:
https://matt.might.net/articles/phd-school-in-pictures/
It's just the fact that there's so much to know that any one person only has the ability to know a tiny, tiny slice of it. Anything else looks alien. You could figure it out with time put into it. But you have to allocate your time, in whatever way seems best.
It's not a bad idea for you to put some of your own time into making good presentations. It's a tricky skill to learn, since it's subjective, but at core it's about empathy -- imaging how the world looks from somebody else's perspective. That's a good talent to have, no matter your field.
I haven't built the kind of tooling he's described but I've used it professionally (Tensorflow, PyTorch) and thought a lot about what's under the hood.
For that kind of math though there is no substitution for doing a good amount of it with your own hands and I paid my dues.
On the other hand I even though I am "just" a coder with a PhD I used to get most of the Springer-Verlag CS conference proceedings from the new book rack at the library and read them on the bus or on the toilet until they shut down the engineering library and quit buying new physical volumes.