Multimodal: a twiddler that combines image and text signals
Only result-level signals are known. RESONEO says it merges more than one kind of input, for example picture and text; which model and with what effect is not documented.
01What is known
The only schema object is MultimodalTwiddlerRanklabResultSignals, signals per result. There is no query object in the list. "Ranklab" appears in the object names; no source available to us explains what that part of the name refers to.
RESONEO describes the twiddler as merging more than one kind of input, for example the picture itself and the words around it. Like every twiddler, it only recommends; the framework in Superroot, the layer that merges results from Google's indexes before the page is built, makes the call.
02What is not known
Which model produces the combined signal, whether it is related to the Starburst image embeddings named in the leak, and how it changes a result's position are all unknown.
On our map it belongs to the twiddler layer and to no E-E-A-T system. It is a rule for how image results are assembled, not a judgement on who made a page.