Well, no-one else has replied, so here goes......NOTE! I'm not an electronics expert nor an aircraft engineer.
A resistance-meter normally works by massing a battery's voltage through it's meter and calibrated resistances,,,,the current through the circuit is thus proportional to the unknown resistance placed on series in the "circuit" the Ua 's of current through the meter are read off the appropriate Ohms scale.
A thermistor alters it's resistance with temperature....I would expect to see the same resistance irrespective of test-meter polarity, though I suppose it's possible that meter-current could heat the thermistor, thus altering it's value dynamically whilst measuring.
Now the interesting bit!.....you talk of a bi-metallic junction device......well, from my limited knowledge, this is akin to a thermocouple or even a semiconductor (transistor, diode? )
One effect of these junctions is that electricity is produced when some are heated (exactly how a thermocouple works)....so , logically, your multimeter's internal battery would ADD with the generated voltage in one connection, and in the opposite connection, they would oppose, therefore, less current would flow in the latter circumstance and the meter would show a high resistance compared with the former situation, where a properly zeroed meter could be expected to show "negative resistance..
explanation....when you join the probes and zero a meter, the current is set by the trimming resistor to read "0 Ohms) (full scale current) when the item tested has voltage, and it's added to the internal battery, the meter is overdriven and "pegs"
Of course, i'm talking proper analogue meters here, but AFAIK
, it's similar with these newfangled digital ones.
Incidentally, the above effect is used to do a quick, rough check on electrolytic capacitors connect,-needle kicks and slowly sinks as cap charges......reverse connection and meter kicks violently and slowly sinks as charge in cap dissipates.
Now wait whilst the professionals tear me to shreds and flay me alive