It depends.

Metabolic and cardiovascular strain (and hence adaptation) will be (primarily) determined by the (temporally-smoothed) power output, regardless of how it is generated.

Neuromuscular demands (and hence adaptations in muscle contractile function) will be (mostly) dictated by the force and speed with which you pedal.

I developed quadrant analysis (QA) about 15 y ago as a way of visualizing the latter demands. QA was inspired, in part, by Robert's interest in torque vs. cadence plots and "power expansion pathways", but explicitly incorporates the physiological knowledge that significant type II motor unit recruitment seems to (initially) occur only at intensities above FTP. (The other motivation for QA was people's complaints that NP didn't seem to fully capture the stress they experienced during severely "on-off" cycling, e.g., a criterium).

Once dual-sided power meters that offered additional information about pedaling dynamics became available, I developed additional analytical tools, i.e., calculation of maximum effective pedal force (MEPF), gross power released (GPR), gross power absorbed (GPA), and kurtotic index (KI), all of which can be found in WKO4. This seems to be a bit of a deeper dive into cycling biomechanics than what motivated your original question, however.

A description of QA can be found in our book, as well as here (note that both the date on the article and the byline are incorrect):

https://www.trainingpeaks.com/blog/quadrant-analysis/

These sources describe the pedaling metrics unique to WKO4:

https://www.trainingpeaks.com/...and-metrics-in-wko4/

https://help.trainingpeaks.com/...ling-metrics-in-WKO4

https://irp-cdn.multiscreensite.com/...daling%20Metrics.pdf

TL, DR answer: specificity, specificity, specificity, specificity, specificity.