Pass a current through them high enough to heat them in air and measure the resistance. That's empty.
As the fuel level rises, the thermistors are cooled by the fuel and the resistance changes.
That would require only two wires to penetrate the tank, so the existing ones can be re-purposed.
- inductance of the spring in the oem sensor might be feasible. TI has a chip made for the job.
- capacitance of 2 plates that extend top to bottom. The dielectric constant of fuel is double that of air so the capacitance changes with fuel level. The only question is the variation of the dielectric constant of different fuels - E10, 98RON, 95RON etc.
A FET to drive the existing gauge - adjusting the gate voltage & monitoring V & I to control the R.
Arduino or Pi for development, but the final processor should be a uC. Way smaller and more easily sealed in a housing. It's built-in supply should be protected against load dumps and reverse battery connection.
There wasn't much interest in a sensor that worked, so I haven't done anything about it. I figured I'd do it when mine died - again.