We examine the performance of standard pre–main-sequence ( PMS ) stellar evolution models against the accurately measured properties of a benchmark sample of 26 PMS stars in 13 eclipsing binary ( EB ) systems having masses 0.04–4.0 M _ { \odot } and nominal ages \approx 1–20 Myr .
We provide a definitive compilation of all fundamental properties for the EBs , with a careful and consistent reassessment of observational uncertainties .
We also provide a definitive compilation of the various PMS model sets , including physical ingredients and limits of applicability .
No set of model isochrones is able to successfully reproduce all of the measured properties of all of the EBs .
In the H-R diagram , the masses inferred for the individual stars by the models are accurate to better than 10 % at \gtrsim 1 M _ { \odot } , but below 1 M _ { \odot } they are discrepant by 50–100 % .
Adjusting the observed radii and temperatures using empirical relations for the effects of magnetic activity helps to resolve the discrepancies in a few cases , but fails as a general solution .
We find evidence that the failure of the models to match the data is linked to the triples in the EB sample ; at least half of the EBs possess tertiary companions .
Excluding the triples , the models reproduce the stellar masses to better than \sim 10 % in the H-R diagram , down to 0.5 M _ { \odot } , below which the current sample is fully contaminated by tertiaries .
We consider several mechanisms by which a tertiary might cause changes in the EB properties and thus corrupt the agreement with stellar model predictions .
We show that the energies of the tertiary orbits are comparable to that needed to potentially explain the scatter in the EB properties through injection of heat , perhaps involving tidal interaction .
It seems from the evidence at hand that this mechanism , however it operates in detail , has more influence on the surface properties of the stars than on their internal structure , as the lithium abundances are broadly in good agreement with model predictions .
The EBs that are members of young clusters appear individually coeval to within 20 % , but collectively show an apparent age spread of \sim 50 % , suggesting true age spreads in young clusters .
However , this apparent spread in the EB ages may also be the result of scatter in the EB properties induced by tertiaries .