Sadly it's not something you're likely to see anymore. I don't think there was ever any obligation to provide it, aside from possible alternate uses for the earliest ICs (like, knowing a CA3028 isn't just a diffamp, but three transistors that can be used as a mixer and other things).
Back in the day, it might've been a matter of pride to display the internal schematic, or at least a useful equivalent to it. Tek and HP had this mentality. Somewhere in the 80s, everyone cranked back on that, preferring to hide the IP, and capitalize it; and perform service by replacing whole boards rather than fix components.
Some have even gone needlessly backwards. Take this for instance:
http://www.diodes.com/_files/datasheets/ZXGD3004E6.pdf It's two fucking transistors. But they never say that that's what it is (even though the test circuit gives it away!). So the risetimes for example are representative of the source, not the device itself...
Even some pin equivalents would be nice, which is frequently missing from analog and mixed signal ICs where it's important. Some National parts did that until relatively recently, I think. TI at large seems here and there, probably depending on who (or which department) was writing it that day...
You can at least take some confident guesses, when you know how ICs are usually made. CMOS logic inputs have protection diodes (or zener, for "5V tolerant" and such types) and low leakage, outputs are complementary ('protection' diodes are implicit, because of body diodes), and pull to the rails with typical strength; etc. But there are plenty of examples of bizarre crap that designers have thrown in there, ranging from analog levels (for e.g. startup conditions, I2C addresses, etc.), to badly designed state machines and completely undocumented features.
So, for R2R amps, you can read into the graphs a little, if the right ones are provided -- if Vos changes at high Vcm, it's some kind of complementary input stage. (What you're seeing is, one of the diff pairs turns off near the rail (which normally results in phase reversal, in poor designs), and the other one isn't as well trimmed so |Vos| goes up.) If it's fairly noisy (~20nV/rtHz), it's probably CMOS (MOSFETs for the input diff pair, at least, but probably the rest as well). If it's bipolar, input bias current will be fairly high (nA+?). Beware of changing (or unspecified!) fT versus supply voltage, Vcm and temp, stuff like that.
Tim