I can imagine the way how the graphs $2$ and $3$can be correct - but this requires the marks on the $y$ axis to not be equidistant, and actually to be very precisely tailored to the picture you want to show.
For example, if on $y$ axis the $1/4$ mark was half-way through (rather than a quarter of the way through) between the $0$ and $1$ marks, then the point $(1/2, 1/4)$ would appear on the straight line on the graph $2$.
Having said all that, unless making up a trick question - why would anyone do that? People expect equidistant marks on both $x$ and $y$ axis, and if that is the case, only graph $1$ is correct. Any other setup of the axes (including the case when the scale is logarithmic on one or both axes) needs to be specifically noted under the graph, so that people would know what they are looking at.
I can imagine the way how the graphs $2$ and $3$ can be correct - but this requires the marks on the $y$ axis to not be equidistant, and actually to be very precisely tailored to the picture you want to show.
For example, if on $y$ axis the $1/4$ mark was half-way through (rather than a quarter of the way through) between the $0$ and $1$ marks, then the point $(1/2, 1/4)$ would appear on the straight line on the graph $2$.
Having said all that, unless making up a trick question - why would anyone do that? People expect equidistant marks on both $x$ and $y$ axis, and if that is the case, only graph $1$ is correct. Any other setup of the axes (including the case when the scale is logarithmic on one or both axes) needs to be specifically noted under the graph, so that people would know what they are looking at.