Is there a paper which first introduced minibatching?
This seems like a tricky question to answer as mini-batch optimisation seems like a natural idea and it might be called something different in different communities. The earliest reference I could easily find is the following:
Bertsekas, Dimitri P. "Incremental least squares methods and the extended Kalman filter." SIAM Journal on Optimization 6.3 (1996): 807-822.
But I would be surprised if the idea wasn't used before then, given the Robbins and Monro paper on stochastic gradient descent came out in 1951.